Hermes Agent

Hermes Agent v0.22.1

October 15, 2026

• •Agent): add nemotron super 3.5 default parameters (`c59d2994`) — @elisam0
• •Cli): add curated model list at the top in the model picker list (`e6835fd3`) — @elisam0
• •Plugin-catalog): add hermes-lang-ko — Korean Desktop language pack (`e5f49b6a`) — @dandacompany
• •Plugin-catalog: add hermes-lang-pt-br (Brazilian Portuguese language pack) (`0169e80c`) — @StudioCodeAI
• •Plugin-catalog: add teamorouter (TeamoRouter model providers) (`ef223406`) — @2867297362mjb-sketch
• •Plugin-catalog): add author-feed (`5f6d6836`) — @TakeshiTGAL
• •Plugin-catalog): bump aux-ledger to 1.4.0 (`6048066e`) — @olric3015
• •Plugin-catalog): add the wait-ledger catalog card (`77cb58e5`) — @olric3015
• •Plugin-catalog): add wait-ledger (`97e65322`) — @olric3015
• •Community plugin: lightningai + lightningai-rate-limit (`07515466`) — @dinosauria123
• •Catalog: add Hermes AAC 0.1.0 at reviewed exact source pin (`7c9adda8`) — @TREE-Ind
• •Plugin-catalog): add openinvest (`b6e86101`) — @longsizhuo
• •Catalog: add hermes-agent-mixer (v0.2.0 @ 04c4e13) (`2e7e3859`) — @Sahil-SS9
• •Plugins): move Kumiho Memory to memory catalog shelf (`46f95e08`) — @kaveone
• •Add canonical group attachment controls (`4673f17c`)

Hermes Agent v0.22.0 — The Threshold Release

October 14, 2026

• •A first launch that sets itself up — Open the desktop app on a computer where you have never used Hermes and it starts a short setup chat instead of a settings screen. One card at a time it asks your name, an accent color and a layout, which apps to connect, which plugins to install, then gives a quick tour and offers up to four first tasks picked for you. It runs on a free trial model (nous/welcome, once per OS user), so it needs no key; Skip setup ends it and Settings → Run onboarding again brings it back. The agent also gained set_theme, so "make it violet, dark mode" works in any desktop chat. (#137429, #137522, #137807, #134209, #105258 — @alt-glitch, @OutThisLife, @ethernet8023, @rewbs)
• •Connectors: Hermes works inside your apps — The agent can now act in hosted apps such as Linear and Notion through the Nous tool gateway: ask for an app, sign in once in your browser, and the tools are usable in the same turn. You do not need a Nous account to start; the first time you use an app, Hermes creates a free connectors-only identity. The desktop has a new Connectors page that puts hosted apps and your MCP servers in one list (it replaces the MCP tab), the same setup card appears in the TUI and CLI, and you can connect several named accounts of one app (a work and a home account) and manage them with /connectors or hermes connectors. (#106842, #111008, #119074, #135143, #135072, #135097, #135151, #135173, #137426 — @alt-glitch, @rewbs)
• •Local models in one click — Hermes can now run open models entirely on your machine with nothing leaving it: no account, no API key. It installs and maintains llama.cpp with the right engine for your hardware (CPU, CUDA, Metal, Vulkan or ROCm), rates every catalog model against your GPU and RAM, and picks the highest-quality build that fits. Downloads pause and resume, the server starts and stops with Hermes, and engine updates are one click. Start from Settings → Providers → Local Models on Windows, macOS and Linux, or hermes model → Local models in a terminal. The catalog includes Qwen3.8 27B, Qwen3.8 Flash Next (with its MTP speculative head), Qwen3.6 35B-A3B and DeepSeek V4 Flash, plus any GGUF from Hugging Face. (#100667, #101823, #128768, #111423, #133111, #134116 — @jquesnelle, @ethernet8023)
• •Stable release channels — Hermes now ships tested SemVer releases. Every release passes source CI, Docker, Nix, signed desktop builds and install/upgrade runs on every OS before it is published, and source installs now move from release to release by default instead of tracking every commit on main. Prefer the bleeding edge? hermes update --set-channel main, or the new channel switch in Desktop Settings → About. Moving to an older release is labelled a rollback and never happens without asking, and installer reruns keep you on the channel you chose. (#102765, #135847, #135949, #136313, #137780 — @ethernet8023, @KoNit-K)
• •Signed desktop packages, the Microsoft Store, and one package manager — The desktop app is now a sealed, signed package that carries its own Python and Node and updates itself, so first launch builds nothing. Windows gets a signed MSIX (x64 and ARM64) through App Installer or the Microsoft Store; macOS gets a signed, notarized build for Apple Silicon. Underneath, a new unified package manager (hermes pm) owns every pinned, hash-checked tool and Python environment: updates build a new environment and switch only once it passes its check, so a failed or killed update can no longer leave you half-applied. (#102765, #132361, #132386, #120339, #122221 — @ethernet8023)
• •The Plugin Catalog: 600 reviewed plugins — The plugin catalog launched this cycle and grew to 600 entries, each pinned to a reviewed commit and installable from Hermes Desktop or the CLI. Browse by category and star ranking with a page per plugin and per author; catalog rules keep listed plugins from patching Hermes internals or shipping unscannable binaries. Memory providers now live with their makers (Honcho, Supermemory, Hindsight, Mem0, OpenViking, Holographic), as do Home Assistant, Spotify and Langfuse: existing users are migrated automatically, and the default install got leaner. (#69446, #111415, #117808, #113381, #114732, #131267, #132476, #136436 — @itpartypattaya, @happy5318, and 400+ plugin authors)
• •The Claude 5.5 and GPT-6 model wave — Claude Fable 5.1, Opus 5.5, Sonnet 5.5 and Haiku 5.5, GPT-6 Astra, Sol and Luna plus GPT-6.1 Sol, and Gemini 4 Argon are pickable on Nous Portal, OpenRouter and native routes where offered. Prior reasoning is now replayed by default on every provider that can read it, across resume, branches and compaction, so long multi-step work keeps its train of thought. And when a provider has a bad hour, Hermes waits out transient 503/529/timeouts in visible, bounded cycles instead of ending your turn. (#100577, #119304, #126842, #134759, #128248, #137955, #136358, #116353 — @kshitijk4poor, @unsupportedpastels, @100yenadmin, @mssteuer, @xxxigm)
• •Sign in to websites with your password manager — The browser agent can log into sites with logins from 1Password, Bitwarden or Hermes's own encrypted vault, and the model never sees the password. If nothing is saved, you get a masked prompt in the CLI, TUI or Desktop to save it, one-time codes are filled without entering the conversation, and card and address fills at checkout need your confirmation. (#106480, #107585, #137084)
• •Bot Screen: watch your bot work, take over when it needs you — A bot running on a Linux gateway can have its own desktop that Hermes Desktop streams live. When it hits a login, a 2FA prompt or a CAPTCHA, you take the screen, finish the step yourself, and hand control back while the bot carries on with your session. With a Docker, SSH or Singularity backend, the bot's screen, computer use and the browser run inside the sandbox, never on your host. (#108914, #121169, #121110)
• •One backend and one gateway for every profile — Hermes now runs a single messaging gateway and a single desktop backend per machine that serve all of your profiles, instead of one process per profile. Creating a profile makes its bots live without a restart, each profile keeps its own credentials, allowlists, cron jobs and sessions, and hermes update migrates existing per-profile setups automatically. Fewer processes, no token clashes, one thing to start and stop. (#112854, #108928, #118273, #117983, #118246, #118027)
• •See and steer your subagents live — Running subagents now stay visible above the composer in the CLI (F6), TUI (Ctrl+T) and Desktop with each child's task, elapsed time and live transcript, plus Steer and Stop buttons. /stop reaches background children and hands back their partial work, and delegate_task can pass screenshots and mocks to vision-capable children. (#105593, #115616, #97689)
• •A model picker that knows your plans — Star your favorite models, see a subscription running low ("17% left · resets 3:00 PM") and when a rate limit lifts right in the picker, and pick reasoning effort as part of choosing a model in hermes model, /model (/model X --reasoning high) and Desktop. You switch before you hit the wall instead of after. (#126535, #133392, #133408, #110346, #113117 — @OutThisLife, @anotherchu)
• •Hermes in your language, everywhere — Every user-facing string in the CLI, messaging gateway, TUI and Desktop now goes through one translation catalog with full coverage in all 16 bundled languages. New languages can ship as plugins, and any profile can override individual strings with a small YAML file. (#126296)
• •Leaner in every sense — The whole codebase was simplified with zero behavior change: source went from about 1.06M to 698K lines (−34%) and gateway/run.py shrank from 34,847 lines to 5,512. Prompts got lighter too: about 19 rarely used tools now load only when the agent searches for them, roughly halving the desktop tool schema sent on every call, and another ~2.1k tokens came off the fixed prompt. Local models prefill less on every turn. (#102117, #97979, #106676, #137084 — @alt-glitch)
• •Small comforts that add up — Opt-in vim keys in the CLI prompt, a git-branch field in the status bar, a bell and desktop notification when Hermes is waiting on you, Ctrl+C that clears your draft instead of killing the turn, an editable /queue, and hermes --native to let your own terminal handle TUI scrollback and search. On Desktop, a new Simple mode turns the app into a plain chat window, and errors now say what went wrong with a one-click fix. (#106412, #106434, #101284, #101320, #137065, #95999, #121035, #118587, #111601 — @saforem2, @TurgutKural, @glitchbunny0, @harshmoney123, @pierrenode, @abhinav11082001-stack, @OutThisLife)

Hermes Agent v0.21.7

October 12, 2026

• •Updates now follow tagged releases. A source install of the official repository updates to the latest published vX.Y.Z release instead of every commit on main. An install already newer than the latest release stays where it is and waits for the next one; nothing is moved back. To keep following every commit: hermes update --set-channel main. Forks and mirrors keep following main. Store, MSIX and bundled desktop installs are unchanged.
• •Choose the channel in the desktop app. Settings ▸ About ▸ Updates has an Update channel switch (Stable releases / Every commit) for source installs, and the update dialog now shows which channel you are on.
• •Reasoning survives resume. Signed thinking replays after hermes chat --resume and on the next gateway turn, and Bedrock Converse reasoning models no longer get rejected on resumed sessions.
• •Updates no longer fail behind some firewalls. Installs on main no longer fetch a release record that regional firewalls could answer with 403. The stable check uses your GitHub token when one is set (GITHUB_TOKEN or gh auth login) and reports a GitHub rate limit as a rate limit.
• •Desktop app: updates itself. Microsoft Store installs update through the Store.
• •CLI: hermes update, or re-run the installer one-liner.
• •Docker / Hermes Cloud: nousresearch/hermes-agent:stable (or :latest).

Hermes Agent v0.21.6

October 8, 2026

# Hermes Agent v0.21.6 Release Date: October 8, 2026 > Patch release. This tag rolls up the ~2,100 PRs merged since v0.21.5 into a stable tagged release for Docker and Hermes Cloud. Full curated notes for this window ship with v0.22.0. <!-- HERMES_BUILDS_TABLE --> ## About this release This is the first release cut by the new stable release pipeline: one attempt ref, a tested Docker image, and a receipt tag at publish. It ships the tag, this GitHub release and the Docker image. The desktop app, Termux packages and the Microsoft Store stay on their current build for now and move with the next bundled release. Measured at commit 818c13be1dc4fd28987e1e881a9408224afd4535, the window since v0.21.5 contains 8,867 non-merge commits across 8,342 changed files (+772,490 / −225,600), 2,106 merged PRs and 3,027 closed issues. Also in the window, undocumented here on purpose: live transcription while you speak on CLI, TUI and Desktop (stt.streaming); spoken voice turns routed to their own model (auxiliary.voice_chat) with reasoning off by default; third-party plugins running in a per-profile plugin host (plugins.isolation: host) and one Settings ▸ Plugins home for their settings pages; favorite and pinned models at the top of the Desktop picker, a usage chip before a subscription hits its wall, and rate-limited providers that say why and until when; local models in hermes model plus llama.cpp CUDA on Linux x64 and arm64; pluggable, layered language packs across core, Desktop and TUI; a summary-first hermes status (--short, --full); a review-pane diff scope selector (Uncommitted / Branch / Last turn) and keep-awake that follows the turn; cron, doctor and home-channel warnings when the cron store goes unwritable; approval flags for credential uploads and invisible Unicode in terminal commands; a Discord setup that checks the bot token and prints the invite link; a unified Kanban workflow definition; activate.fish; GPT-6.1 Sol and Claude Sonnet 5.5 / Haiku 5.5 in the catalogs; and dozens of new community plugins. Full curated release notes for this window ship with v0.22.0, which will document everything from v0.21.0 onward — highlights, feature areas, and complete contributor credits. Nothing in this window is skipped. ## Updating - Docker / Hermes Cloud: nousresearch/hermes-agent:stable (or :latest). - CLI (git installs): hermes update, or re-run the installer one-liner. - Desktop app and Termux: no change in this release; they update with the next bundled release. ## Security Fixes Dashboard authentication hardening - Spoofed X-Forwarded-For headers could reset the password-login rate limit and get around the per-IP cap on native sign-in (#133367) - Unauthenticated login requests could write unbounded values to the auth audit log (#133369) - The public /auth/ routes had no request-body size limit (#133370) - Native sign-in could send login codes to a non-loopback redirect, which allowed session takeover (#130685) Thanks to Tenable Research for reporting all four issues (TRA-725, TRA-726, TRA-727, TRA-728). The native sign-in redirect issue was also reported independently and fixed by @Froraut. The X-Forwarded-For fix builds on community work by @hinotoi-agent and @BearHuddleston (#40285). Repository git filter hardening - Automatic git calls (session workspace snapshot, subagent worktrees, kanban, worktree cleanup, hermes -w) could run clean/smudge/process filter programs defined by an untrusted repository's own config, before the first prompt (#130661) Thanks to 燕涛 for reporting this privately. It was also reported and fixed publicly by @jonpol01 (#126017, #126019) and @JoaoMarcos44 (#126075). Email gateway sender hardening - A quoted display name in the From header (e.g. "Victim <victim@example.com>" <attacker@evil.test>) could make an attacker's message pass the email allowlist as an allowed sender (#125212) Thanks to Daniel Steele (@keeltrace) for reporting this and writing the fix (#124322). Full Changelog: v2026.9.24...v0.21.6

Hermes Agent v0.21.5 (v2026.9.24)

September 24, 2026

• •hermes update (git installs), or re-run the installer one-liner.
• •Docker / Hermes Cloud: images build from this tag (nousresearch/hermes-agent:v2026.9.24).

Hermes Agent v0.21.4 (v2026.9.21)

September 21, 2026

• •hermes update (git installs), or re-run the installer one-liner.
• •Docker / Hermes Cloud: images build from this tag (nousresearch/hermes-agent:v2026.9.21).

Hermes Agent v0.21.3 (v2026.9.14)

September 14, 2026

• •Remote dashboard sessions no longer expire on refresh bursts (#110061, fixes #55712; salvage #71548 @Doud-FR, #55717 @liuhao1024). Both refresh paths on the gateway (cookie gate and the desktop's native bearer route) now coalesce concurrent requests carrying the same rotating refresh token, so a Desktop wake burst can no longer replay an already-rotated token into the Portal's reuse detection and revoke the whole session. Refresh also runs off the event loop, so a slow identity provider no longer freezes /api/status. Pairs with Portal-side NousResearch/hermes-portal#1209 (sliding 30-day idle horizon, 5-minute rotated-token grace).
• •Long-lived processes stop leaking duplicate state.db writer handles (#110934, fixes #100896 #103339; salvage #107974 @kshitijk4poor, mapping @Rroven): gateway, dashboard/Desktop backend, ACP and CLI readers attach read-only and in-process writers share the registry handle, so the N live SessionDB handles precursor stops firing on a healthy topology.
• •hermes update (git installs), or re-run the installer one-liner.
• •Docker / Hermes Cloud: images build from this tag (nousresearch/hermes-agent:v2026.9.14).

Hermes Agent v0.21.2 (v2026.9.11)

September 11, 2026

• •No more second writers. Profile gateways wrote hosted-room state into the root state.db every 5 seconds; the dashboard opened a writable handle on startup; cron's lifecycle guard did a raw open() on a live database (which cancels the gateway's POSIX locks — the classic "how to corrupt SQLite" recipe); doctor --fix would checkpoint under a live holder. All four are gone: hosted rooms live in shared-state.db, the dashboard opens read-only first, the guard goes through the tracked connection registry, and doctor --fix refuses a checkpoint it can't prove is safe. (#108076 — salvage #103489 @RikETS, #102682 @JoaoMarcos44, #108012 @Halldrix, #105428 @TaoMasterCoder)
• •Healthy WAL databases stop wedging. OpenZFS (deleted) dentries and a close() racing an append_message both produced a sticky DeletedWalGenerationError on a perfectly good store; the read pool was handed out under an unconfirmed journal mode; a transient disk I/O error on WSL2 killed get_session on the first attempt; and a "state.db locked" banner was broadcast after the lock had already cleared. (#108082 — salvage #107411 @chelsealong, #105578 @ca-shrimp, #105711 @gaoanze888, #106958 @nikkoxgonzales; co-authored @QDung210, @fangliquanflq, @Sahilvishnaliya)
• •FTS damage no longer kills your turn. An error scoped to the full-text-search index was classified as whole-file corruption and fail-closed the conversation. It's now fts_index: search degrades, the index rebuilds later, the transcript store is untouched. Same PR: doctor names structural damage honestly instead of "FTS write corruption", the FTS write probe catches the stale-index shape that passed every check while every write failed, .recover output no longer fails startup on orphan FTS5 shadow tables, header-zeroed databases recover instead of being refused, and the dashboard analytics poller returns a 503 instead of 520K tracebacks a day. (#108130 — salvage #97843 @SulthanZahran1 + #97841 @Finn763, #88604 #56824 #103657 @liuhao1024, #106890 @nftpoetrist, #103321 @jangomango76, #91413 @leegunwoo98, #102808 @TaoMasterCoder)
• •One corrupt row no longer kills `sessions list`, export, or insights. A TEXT timestamp or a 1e30 epoch used to crash the whole listing; malformed marker JSON crashed json_extract; more than 999 ids crashed bulk delete/prune. One coerce_epoch() helper on every reader (bad rows render ? with a WARNING naming the session), a json_valid guard, IN-list chunking, and batched export hydration. (#108086 — salvage #106071 @Xipong, #101726 @efe-arv, #94701 @liuhao1024, #102679 @mssteuer, #100658 @Mi55ed)
• •Sessions never bind to or read another profile's database. The Desktop launch backend could pin itself to the wrong profile's state.db under a HERMES_HOME override race; session_search by bare ID silently scanned every profile and returned someone else's transcript; recovery guidance pointed at the wrong file; profile delete kept a handle open (WinError 32). (#108074 — salvage #102534 @HexLab98, #106975 @Sora-bluesky)
• •Opening state.db no longer takes the write lock when nothing needs writing. A one-shot hermes process opening the store behind a busy gateway stalled 4–20 s and then failed with "database is locked". Now 0.01 s. (#108067 — salvage #106751 @kshitijk4poor, #101881 @jonpol01)
• •A bare display: key in config.yaml no longer crashes every gateway turn (#106305); a queued-lane final refused by the platform is recorded and redelivered (#106316); a stalled WebSocket send no longer blocks every later event (#106581); the first turn no longer waits on the Python toolchain probe (#106556).
• •Telegram bots must @mention when bots_require_mention is on, breaking bot-to-bot loops (#106534); Matrix renders LaTeX (#106515); Signal renders markdown tables (#106538); WhatsApp replies to view-once messages keep their quote (#106541); media-only replies report SUCCESS everywhere (#106557).
• •/model and auxiliary auto never bill a provider you didn't select (#107366); never auto-switch to a provider you have no credentials for (#107281); Bedrock Claude/Converse/Mantle models survive /model, fallback and restore (#107621, #107658); Bedrock Guardrails enforced (#107815).
• •Codex: patch-budget image 400 shrinks and retries (#106525); unentitled primary + fallback no longer oscillate (#106549); Azure Foundry replayed-reasoning rejection classified and pruned (#106718 @erosika). MCP OAuth refresh no longer erases the refresh token (#106185). Anthropic clients send exactly one credential (#107978).
• •DeepSeek V4.1 Flash on Nous Portal and OpenRouter pickers (#107489); GPT Image 2.5 via OpenAI and FAL (#105988); Opus 5 / Fable 5.1 on the native Anthropic picker (#106636 @xxxigm).
• •One blocked periodic callback no longer stalls lease refresh (#106308); a mid-turn /steer is persisted as its own user row (#106317, #106344); local-inference memory-ceiling rejections back off instead of compressing history (#106307); context-overflow after partial streaming ends the turn (#106567); length continuation stops when the prompt filled the window (#106571); compression no longer times out silently on aux retries (#106866); model_thresholds keys can be provider-scoped (#108061).
• •Surface switch (Desktop↔TUI) no longer rebuilds the system prompt and busts the prompt cache (#105844); CLI keeps the api_content sidecar so the cache survives an early persist (#105842).
• •hermes -z --resume continues the session (#106313); Shift+letter and Cmd+Shift+Z work on extended-key terminals (#90674 @francip, #105493); browser_exec timeout kills the whole process tree (#106589); update checks poll the GitHub API once a day instead of git-fetching every 30 min (#107648); hermes update names the real cause and can't hang on a stalled fetch (#108053).
• •Desktop: UI language survives the update relaunch (#106476), expired OAuth grants get a one-click re-sign-in (#106965), HUD mode shows the transcript again and always gives the window back (#107491, #107423), the backend exits when its Desktop parent dies (#107977), Windows updates stop reporting false failures (#106175, #107183), WSLg renders on the Windows GPU (#106528), Telegram quick setup with QR ported from the dashboard (#107242), and ~60 more Desktop fixes largely from @OutThisLife and @kshitijk4poor.

Hermes Agent v0.21.1 (v2026.9.7)

September 8, 2026

• •Existing install: hermes update
• •Fresh install: curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
• •Managed deployments should update through their deployment tooling using the new tag.

Hermes Agent v0.21.0 (v2026.8.31)

August 31, 2026

• •Bot Mode — your agents become a society, built in — Bot Mode is now a bundled, default-on part of the desktop app: every agent profile gets a name, a deterministic avatar face (with randomize/lock controls), and a place in a shared roster. Create Discord-style group chats where multiple bots and you talk in one room, @-mention any bot from the composer, and give rooms names and pictures. Before, "multi-agent" meant plumbing; now it looks like a chat app full of coworkers. (#87886, #88243, #89386, #96726 — @teknium1, @OutThisLife, @dokterdok)
• •`hermes peer` — bot-to-bot DMs between your agents — Any Hermes agent can now message any other by handle, across profiles and gateways, from the CLI or from inside a conversation. Ask your research bot to hand findings to your coding bot and get the reply back where you can read it. Replies land in each agent's canonical Bot Chat, so conversations between agents are durable and inspectable, not fire-and-forget. (#88725, #88178, #91487 — @teknium1)
• •Cron jobs that remember — Scheduled jobs stopped being goldfish. Cron agents now load and update persistent memory like every other agent, continuity=true carries each run's output into the next (so a monitor can dedupe against what it already reported), every job gets a durable notepad scratchpad, monitor-mode jobs skip the LLM entirely when nothing changed, and cron output can land in a bot's canonical Bot Chat — where the bot actually responds. Your 9am briefing job now knows what it told you yesterday. (#91447, #80774, #81139, #81138, #91487 — @teknium1, @smwbev)
• •Steer your subagents while they run — delegate_task gained live orchestration: list running children, steer one mid-flight with a course correction, or stop it early and keep the partial result. Add optional JSON-schema validation on child outputs, per-delegation cost surfaced in results, and raised defaults (250 iterations, 10 concurrent children) — delegation went from fire-and-pray to actually managed parallel work. (#85232, #81144, #81142, #86506, #86745 — @teknium1)
• •The MCP command center — MCP servers and the catalog merged into one coherent desktop page with drag-in "paste anything" import, background health checks that nudge you to re-auth before a tool call fails, a fleet cost/usage overlay showing schema token estimates and 30-day usage per server, and hermes:// deep links that install an MCP server with explicit confirmation. Managing twenty MCP servers used to be config-file archaeology; now it's a dashboard. (#87525, #87572, #87576, #87579, #87581 — @teknium1)
• •A CLI power wave — Ctrl+P opens a fuzzy command palette, the /model picker filters as you type, /status shows reasoning mode, pending approvals, and context usage, and the status bar can display live cache-hit %, latency, and tokens/sec with per-field toggles. Plus: a global emergency stop, session pin/unpin, rotating task-oriented composer placeholders — and Ghostty-level terminal pets, because a companion should have a companion. (#90730, #90717, #90745, #98250, #98282, #97666 — @teknium1, and salvaged community work)
• •The agent drives the desktop's browser — The in-app browser stopped being a window the agent could only look at: Hermes now navigates, clicks, and reads it directly, and pages can be popped out to your system browser with full link context menus. Ask it to walk a docs site or debug a web app and watch it happen inside your own app. (#90197, #89366 — @OutThisLife, @ethernet8023)
• •Six new providers and a model catalog wave — Meta Model API (Muse Spark) arrives as a built-in provider, alongside CommandCode, Tencent TokenPlan, Nebius Token Factory, Ramp Router, and Actual Computer. The catalogs picked up GLM-5.3-Flash, qwen3.8-max/flash, Gemini 3.7 Flash, MiniMax M3 free, and Nemotron 3.5 Lightning — and model_overrides lets you patch any model's context window or pricing yourself without waiting on a release. (#88565, #88308, #97917, #97916, #97915, #85560 — @teknium1 and community)
• •Security hardening across the board — Protected agent-instruction files (AGENTS.md, skills, memory stores) now always require write approval so a prompt-injected agent can't quietly rewrite its own standing orders. A deep redaction sweep closed secret-leak gaps across terminal errors, .env file reads, checkpoints, and ACP logs; the approval system learned Windows destructive commands; and macOS permission grants finally survive updates via a stable TCC signing identity. (#81152, #80965, #84428, #95091 — @teknium1 and community)
• •New providers: Actual Computer inference (#79644, salvage of #26491), CommandCode with GOAT/Pro/Max plans (#88308, salvage of #32909), Meta Model API (Muse Spark) as a built-in provider plugin (#88565), Tencent TokenPlan (#97917), Nebius Token Factory (#97916), and Ramp Router (#97915).
• •Model catalog wave: qwen3.8-max and qwen3.8-flash (#78024, #96979), Gemini 3.7 Flash (#85526), GLM-5.3-Flash across OpenRouter/Nous/z.ai/OpenCode (#95621, #96293), MiniMax M3 + Inkling free SKUs (#96264), Nemotron 3.5 Lightning (#84645), Meta Muse Spark 1.2 (#88600).
• •Per-model metadata overrides via model_overrides config — patch context windows, pricing, or capabilities for any model without waiting on a release (#85560).
• •Data-training-tier warnings — a unified selection-guard registry warns you across every picker surface when a model trains on your data (#85917).
• •Pip-installed model providers discovered via entry points — third parties can ship providers as packages (#85504, salvage #81419).
• •Prompt caching engaged for LiteLLM Claude on the OpenAI wire (#87517 — @kshitijk4poor) and api.meta.ai routed through the Responses API for caching (#88601).

Hermes Agent v0.20.6 (v2026.8.27)

August 27, 2026

• •hermes update from an existing install
• •Fresh install: curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Hermes Agent v0.20.5 (v2026.8.19)

August 21, 2026

• •hermes update from an existing install
• •Fresh install: curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Hermes Agent v0.20.4 (2026.8.18)

August 18, 2026

• •hermes update from an existing install
• •Fresh install: curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Hermes Agent v0.20.3 (2026.8.16.2)

August 17, 2026

• •hermes update from an existing install
• •Fresh install: curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Hermes Agent v0.20.2 (2026.8.16)

August 16, 2026

• •hermes update from an existing install
• •Fresh install: curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

Hermes Agent v0.20.1 (2026.8.13)

August 13, 2026

• •Existing installs: hermes update
• •Fresh install: see the installer one-liner in the README

Hermes Agent v0.20.0 (2026.8.3)

August 3, 2026

• •Talk to Hermes — streaming, conversational voice with barge-in — Voice mode used to mean: speak, wait for the whole reply to generate, then listen to one long audio file. Now Hermes speaks clause-by-clause as the response streams, you can interrupt it mid-sentence by just talking (it stops, listens, and the model is told you cut in), and busy-aware silence detection means it doesn't talk over you. This works in CLI voice mode, on the desktop, and through gateway adapters. Talking to Hermes finally feels like a conversation, not a voicemail exchange. (#69511, #73862, #74223, #74000, #69602 — @teknium1, @OutThisLife)
• •Wake words and hands-free control — Say your own open-vocabulary wake phrase ("hey Hermes", or anything you pick) and Hermes starts listening — detection runs on-device, so no audio leaves your machine while it waits. Multi-profile voice routing means different wake words can reach different profiles, and saying "stop" ends the voice chat on every surface without touching the keyboard. Your terminal is now something you can talk to from across the room. (#70509, #73106, #73933 — @teknium1)
• •Voice on every platform — Send a voice note to Hermes on WhatsApp, Feishu, DingTalk, LINE, QQ, Photon, or Weixin and it's transcribed and answered; auto-TTS replies are delivered platform-aware (opus where platforms want opus, captions attached correctly). STT is now fully configurable — its own hermes tools category, GUI toggles, dashboard dropdowns, unified language resolution so transcripts stop coming back in the wrong language, and OpenAI's gpt-transcribe support. One unified spoken-text preprocessor cleans markdown, code, and URLs out of speech across all TTS providers. (#73515, #73508, #73910, #73513, #73067 — @teknium1)
• •Research you can trust — grounded citations with fact-checking — The new grounded-citations skill makes Hermes produce research where every claim is backed by a verifiable source: quotes are matched against the actual page text (not hallucinated), citations link to the exact evidence, and a fact-checking mode turns the same machinery on any document or claim you hand it — it tells you what checks out, what doesn't, and what couldn't be verified. If you use Hermes for research, this is the difference between "sounds right" and "provably sourced." (#71698, #77104 — @teknium1)
• •Outbound webhooks — Hermes pushes events to your systems — Until now, integrating with Hermes meant polling or listening on a platform. Now Hermes pushes signed lifecycle events (session activity, turn completions, tool events) to any HTTP endpoint you register — with HMAC signatures so your receiver can verify authenticity. Wire Hermes into your CI, your home automation, your dashboards, or any service that speaks HTTP, with no polling loop. (#69406 — @teknium1)
• •The desktop app becomes a platform — artifacts, plugin SDK, quick entry — Hermes desktop now renders artifacts: versioned cards with sandboxed live preview in a right-rail viewer, so generated HTML/apps run safely next to the chat. A real plugin SDK landed with Kanban as its founding plugin, ctx.download for handing users files, floating pane placement, and multiple GUI windows. A global-hotkey quick-entry window captures a thought into any session from anywhere in your OS. The desktop stopped being a chat client and started being a workbench. (#72345, #61173, #74413, #72315, #68259, #73143 — @OutThisLife, @teknium1)
• •Hermes speaks Agent-to-Agent — A2A v1.0 — A new bundled plugin implements the Agent-to-Agent protocol, so Hermes can discover, talk to, and be driven by other A2A-compatible agents. This closes issue #514 — one of the oldest open feature requests in the repo. If you're building multi-agent systems with heterogeneous stacks, Hermes now has a standard wire protocol for joining them. (#77109 — @teknium1)
• •CLI power-user wave — !command runs a shell command instantly without spending a model turn. /init scans your project and generates (or updates) an AGENTS.md. /diff shows staged/all/session changes from any surface, /context breaks down exactly what's filling your context window, /focus gives you a reduced-output view with hidden-line recovery, and Ctrl+S stashes a half-written prompt into a browsable panel. Plus hermes import-agent migrates your Claude Code or Codex CLI setup into Hermes in one command. (#72257, #72178, #72240, #72242, #72302, #72262, #72190 — @teknium1, several salvaging long-standing community PRs)
• •Correct the agent mid-turn — redirects — If Hermes is heading the wrong way, you no longer have to /stop and re-explain. Type a correction while it works and the active turn is redirected: work in flight is preserved, the original prompt is kept, and the agent course-corrects with your new guidance. Paired with double-ESC draft discard and a composer undo stack, steering feels like editing, not restarting. (#63104, #72339, #74736 — @OutThisLife)
• •Tools that fix themselves — A sweep of self-recovery upgrades means the agent wastes far fewer turns on tool friction: truncated terminal output spills to a file the agent can read back, patch detects already-applied edits and diagnoses whitespace mismatches, write_file verifies content on disk, searches that match nothing probe for near-misses and recover, and common failure classes come back with actionable hints. The default tool-calling iteration limit also jumped 90 → 500 — long autonomous runs stopped hitting an artificial wall. (#77041, #76998, #77024, #77055, #77011, #76992, #72176 — @teknium1)
• •Compression that respects your conversation — Context compression got a deep overhaul: proactive tool-result pruning for large-window models, per-turn micro-compaction that amortizes the cost instead of one giant pause, a guaranteed N-user-message tail so recent conversation always survives, progress-aware timeouts that stop punishing slow summary models, and ghost-skill defense so a pruned skill can never silently haunt a session. Thresholds are now configurable per-model and in absolute tokens. Long sessions stay coherent and stop stalling. (#70254, #75345, #70250, #71508, #70275 — @teknium1, @kshitijk4poor, salvaging multiple community PRs)
• •Smart approvals grow up — hermes approvals suggest mines your approval history into allowlist proposals, operators can customize the smart-approval policy, a consecutive-denial circuit breaker stops a misbehaving loop cold, and desktop pairing approvals are profile-correct with a proper surface to answer them from. Plus a new approval gate for docker/podman daemon-redirect commands. Less clicking "approve", without giving an inch of control. (#72259, #72186, #72203, #74446, #71092 — @teknium1, @OutThisLife)
• •Faster everywhere, again — Prompt caching now covers tool schemas on native Anthropic without history loss. hermes -w cold start dropped ~14s → ~1.8s, hermes update no-ops got 2–6s faster, heavy SDKs lazy-load off the import path, config reads stopped deep-copying (54× faster on the telemetry gate), and the desktop shipped a second 60fps wave — streaming cost independent of transcript length, drag at 60fps with five streaming tabs, idle CPU near zero in the background. (#76032, #71637, #74218, #74204, #71835, #72346, #75218 — @kshitijk4poor, @teknium1, @OutThisLife)
• •New places to run and be reached — Buzz lands as a bundled gateway platform (Block's Nostr-based messenger, with native WebSocket transport and NIP-42 auth), the Vercel AI Gateway provider and Vercel Sandbox terminal backend return modernized, desktop gains an SSH remote-backend connection mode, and the Relay shipped four phases of parity — media, interactive prompts, thread lifecycle, typing indicators — plus HSP personal + org skill sync. (#73610, #73761, #74518, #68130, #71300–#71624, #66730 — @teknium1, @yoniebans, @benbarclay)
• •Streaming, conversational TTS with barge-in across all surfaces; clause-by-clause synthesis for CLI voice mode + gateway adapters (#69511, #73862 — @OutThisLife, @teknium1)

Hermes Agent v0.19.1 (v2026.7.30)

July 31, 2026

# Hermes Agent v0.19.1 (v2026.7.30) Release Date: July 30, 2026 > Patch release. This tag rolls up the ~1,000+ PRs merged since v0.19.0 into a stable tagged release for downstream consumers (Docker images, hosted deployments, fresh installs). --- ## About this release Since v2026.7.20 (v0.19.0, July 20): ~2,789 commits · ~4,748 files changed · ~442,000 insertions · ~392,300 deletions on main. This window is dominated by bug-fix and salvage waves across the gateway, voice subsystem, desktop app, and installer, plus continued platform work (Buzz/Nostr channel, FLUX3 video generation and delivery, Telegram media reliability, voice-mode regressions). Full curated release notes for this window will ship with v0.20.0, which will document everything from v0.19.0 onward — highlights, feature areas, and complete contributor credits. Nothing in this window is skipped. ## Updating ```bash hermes update # or fresh install: curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash ``` Full Changelog: v2026.7.20...v2026.7.30

Hermes Agent v0.19.0 (2026.7.20) — The Quicksilver Release

July 20, 2026

• •Hermes got dramatically faster — first token in a fraction of the time — Cold-start "Initializing agent..." used to eat ~4.3 seconds before your first turn even reached the model; it's now ~0.9s, an ~80% cut that applies to the CLI, gateway, TUI, desktop, and cron alike. Round 2 attacked what you see while waiting: reasoning models now stream their thinking live by default (no more staring at a spinner for 30 seconds), and the response box paints per token instead of per line. If Hermes ever felt like it took a deep breath before answering, that breath is gone. (#59332, #59389 — @teknium1)
• •The desktop app speed wave — 20+ targeted perf PRs — Long replies used to cost 14× more CPU in the markdown splitter than they do now; giant diffs froze the review pane until we virtualized it; switching sessions thrashes layout no more. Streaming no longer re-renders the sidebar and every tool row per token, profile backends pre-warm on hover intent, and boot-hidden panes mount at idle instead of on the cold-start critical path. The net effect: the desktop app feels like a native app under load, even with huge transcripts and busy agents. (#67154, #67818, #65898, #66033, #66747, #67742 and more — @OutThisLife)
• •Manage your Nous plan from the terminal — `/subscription` and `/topup` — Changing your subscription used to mean a trip to the billing website. Now /subscription opens a full flow right in the TUI or classic CLI: see your plan and remaining allowance, preview exactly what an upgrade costs ("Pay $46.30 & upgrade now") or when a downgrade takes effect, and apply it — with scheduled-change banners and undo. The desktop app got a matching billing settings tab. Your wallet never has to leave the keyboard. (#51639, #61054, #61067 — @alt-glitch)
• •Smart approvals are now the default — When Hermes wants to run a flagged command, an LLM reviewer now assesses it independently instead of asking you to approve every single one — and each verdict covers only that exact command, so a later command matching the same pattern gets its own review. Combined with the new user-defined deny rules (which block commands even under yolo mode) and /deny <reason> (which tells the agent why you refused so it course-corrects), day-to-day approval fatigue drops sharply without giving up control. (#62661, #59164, #54518 — @teknium1)
• •Plug your password manager into Hermes — Bitwarden & 1Password secret sources — API keys no longer have to live in a plaintext .env. A new pluggable SecretSource interface lets Hermes fetch secrets from Bitwarden and 1Password (op:// references) at load time, with multiple vaults enabled simultaneously, deterministic precedence, conflict warnings, and per-variable provenance. This consolidated eleven competing community PRs into one orchestrated interface — future vault providers drop in as plugins. (#59498 — @teknium1, 1Password provider salvaged from @hwrdprkns)
• •Watch your subagents work — live transcripts + durable background delegation — delegate_task dispatches now return live transcript files you can tail -f the moment the subagents launch: every tool call, result, and streamed reply, one human-readable log per child. And background delegation completions are now durable — if the process restarts mid-run, results are restored and delivered through an ownership-checked ledger instead of vanishing. Fan out a fleet, watch any worker live, and never lose the results. (#67479, #63494 — @teknium1)
• •A finished answer can no longer be lost — the delivery-obligation ledger — If the gateway died between generating your response and confirming the platform actually delivered it, that answer used to be silently gone (and you'd paid for the turn). Final responses are now recorded in a durable ledger in state.db around the platform send and redelivered on the next boot — closing a P1 silent-loss window for Telegram, Discord, Slack, and every other channel. (#67181 — @teknium1)
• •One gateway, many profiles — profile-based message routing — A single multiplexed gateway sharing one bot token can now route specific guilds, channels, or threads to different profiles — each with fully isolated config, skills, memory, and secrets. Point your work Discord server at the work profile and your hobby server at personal, from one bot. A second multiplex hardening wave means one misconfigured profile can no longer take down the whole gateway. (#64835 salvaging @Burgunthy, #65700, #60589 — @teknium1, @benbarclay + six salvaged contributors)
• •New providers and the newest frontier models — Fireworks AI and DeepInfra land as first-class providers (Fireworks with cost estimation and a #2 slot in the provider picker), Upstage Solar joins via salvage, and the model catalogs picked up GPT-5.6 (Sol/Terra/Luna + Pro variants, wired end-to-end across every route), grok-4.5 (GA), moonshotai/kimi-k3, claude-fable-5 / claude-sonnet-5, and GA tencent/hy3 — plus LM Studio JIT model loading for local setups. (#62593, #63969, #61616 — @kshitijk4poor completing @rob-maron's #61578, #60887, #65913, #64541, #65472)
• •Crank the thinking to max — new reasoning effort tiers and per-model control — Reasoning effort gained max and ultra levels (GPT-5.6 and Codex's top tiers), selectable everywhere from the CLI to the desktop, with sane clamping on providers with smaller scales. You can now also pin per-model reasoning-effort overrides in config, set per-slot effort in MoA presets (your advisors think hard, your synthesizer stays fast), and per-task effort for auxiliary models. Thinking depth is now a dial, not a global switch. (#62650, #64458, #64631, #64597 — @teknium1)
• •Your sessions, your data — export everything — hermes sessions export now writes Markdown, Quarto, HTML, prompt-only, and even Hugging Face-ready trace formats, with the full filter surface (age, workspace, platform), an opt-in --redact secret-scrubbing pass, and compacted-session lineage stitched into one logical export. Pair with the new prune filters and bulk archive to keep your session store tidy. Your conversation history is a real dataset now, not a black box. (#60186 salvaging @web3blind, #60492, #60507, #59327 — @teknium1)
• •Security hardening round — This window closed a long list of credential-surface gaps: Vertex credentials scoped away from subprocess env and through profile secret scopes, media/vision/image-gen local-file reads routed through one shared credential-read guard, a webhook body-size-cap sweep across every aiohttp server, bot-token redaction in Telegram transport errors, Fireworks token prefixes added to the redactor, six P1 browser/MEDIA/.env hardening PRs salvaged in one pass, and CI hardened against untrusted-ref interpolation. (#57660, #58709, #59215, #56582, #57842 — @teknium1, @srojk34, @kshitijk4poor, @jquesnelle)
• •~80% TTFT cut — Discord capability detection off the critical path (token-keyed 24h disk cache + background refresh), Ollama probe skipped for known non-Ollama providers, agent-init blocking work removed; cold submit→dispatch ~4.3s → ~0.9s (#59332 — @teknium1)
• •Perceived-latency round 2 — display.show_reasoning default ON (watch the model think instead of a spinner), per-token response-box painting with width-aware force-flush, prompt-build caching, mtime-cached timezone resolution (#59389 — @teknium1)
• •Segment mixed tool batches to recover lost concurrency; drop per-call base64 re-serialization from request-size estimates (#64460, #67788 — @teknium1, @OutThisLife)

Hermes Agent v0.18.2 (2026.7.7.2)

July 8, 2026

• •fix(whatsapp): unpin Baileys from git commit, use published 7.0.0-rc13 (#60643) — the WhatsApp bridge dependency now installs from the published npm release instead of a pinned git commit, making installs and Docker image builds reliable.

Hermes Agent v0.18.1 (2026.7.7)

July 8, 2026

# Hermes Agent v0.18.1 (v2026.7.7) Release Date: July 7, 2026 > Patch release. This tag rolls up the ~660 PRs merged since v0.18.0 (July 1) — bug fixes, hardening, and in-progress feature work — into a stable tagged release for downstream consumers (Docker images, hosted deployments, PyPI installs). --- ## About this release This is an infrastructure-driven patch tag rather than a fully curated release. Since v0.18.0 shipped six days ago, main has accumulated roughly 667 commits across ~990 files (+89.5k/−10.4k lines), including installer/updater self-healing on Windows, dashboard and gateway fixes, WhatsApp dashboard pairing, MCP and provider fixes, and a large volume of stability work. Full curated release notes for this window will ship with v0.19.0, which will document everything from v0.18.0 onward — highlights, feature areas, and complete contributor credits. Nothing in this window is skipped; it's documented in the next minor release. ## Updating ```bash hermes update # existing installs pip install -U hermes-agent ``` Full Changelog: v2026.7.1...v2026.7.7

Hermes Agent v0.18.0 (2026.7.1) — The Judgment Release

July 1, 2026

• •Mixture-of-Agents is now a first-class model you can pick — MoA used to be a mode you toggled; now every named MoA preset shows up as a selectable model under a moa provider, right alongside Claude, GPT, and Grok in every model picker (CLI, TUI, desktop, gateway). Pick "my-council" the same way you'd pick any model, and Hermes routes your prompt through that ensemble automatically. An ensemble of frontier models deliberating on your hardest questions is now one selection away, on every surface. (#46081, #53548, #53561 — @teknium1)
• •See every model's reasoning, then watch the answer stream in — When a MoA ensemble runs, each reference model's full output now renders as its own labelled block — you can read what GPT-5 thought, what Claude thought, and what Grok thought, before the aggregator synthesizes them into one answer. And that final answer now streams to you live instead of appearing all at once after a long silence. This works in the CLI, the TUI, and the desktop app. You get to watch the committee deliberate, not just read the verdict. (#53793, #53855, #55625, #56101 — @teknium1)
• •The agent verifies its own work — "done" means proven, not claimed — Hermes now records verification evidence for coding work and can decide it's finished by actually running your project's checks, not by asserting success. /goal gained completion contracts: you state what "done" looks like, and the standing-goal loop judges completion against that evidence instead of stopping when the model feels like it. There's a pre_verify hook for wiring in custom checks and a one-time migration that tunes the defaults sensibly. The difference between "I think I fixed it" and "the tests pass, here's proof." (#50501, #52285, #55413, #53552 — @teknium1, @OutThisLife)
• •`/learn` — turn anything into a reusable skill by describing it — Run /learn <anything> and Hermes distills a reusable skill out of whatever you point it at — a directory, a URL, or just the workflow you walked it through five minutes ago. It writes the skill to the standards in your CONTRIBUTING.md automatically. The next time you need that workflow, it's already there. Teaching Hermes a new trick is now a single command, not a manual skill-authoring session. (#51506, #52372 — @teknium1)
• •`/journey` — a playable timeline of everything Hermes has learned about you — The CLI and TUI gained /journey, a learning timeline that shows the memories and skills Hermes has accumulated over time — and you can edit or delete any of them right from the view. Pair it with the desktop's new memory graph (a top-down, playable radial timeline of memories and skills) and for the first time you can actually see what your agent knows, watch it grow, and prune what's wrong. Your agent's memory stops being a black box. (#55555, #55859, #55226 — @OutThisLife)
• •Delegate a pile of work and keep going — background fan-out — delegate_task can now fan out multiple subagents that all run in the background: your chat is never blocked, and when every subagent finishes, their results come back as a single consolidated turn. Kick off "research these five competitors in parallel" or "audit these three modules," then carry on with something else while a small fleet works. When it's all done, you get one clean summary instead of babysitting each one. (#49734 — @teknium1)
• •First-class coding Projects in the desktop app — The desktop app gained real, per-profile Projects — a sidebar of your codebases, a coding rail, a review pane, git worktree management, and agent-facing project tools, all backed by a proper project → repo → lane model. Instead of scattered chat sessions, your coding work is organized into projects the agent understands and can act on. It's the desktop turning into an actual coding cockpit. (#49037, #54385, #54517 — @OutThisLife)
• •Run Hermes at scale — scale-to-zero and drain coordination — The gateway can now go dormant when idle and quiesce cleanly before a restart, migration, or auto-update — without dropping in-flight conversations. A hosted or relay-only Hermes can scale to zero when nobody's talking to it and wake back up on demand, and disruptive lifecycle actions coordinate an external drain so nobody gets cut off mid-turn. Running Hermes for a team or as a hosted service just got a lot more production-grade. (#52243, #52937, #54824 — @teknium1, @benbarclay)
• •Cheaper self-improvement — smarter background review — The post-turn self-improvement fork (the one that decides whether to save a memory or skill) now routes to an auxiliary model, digests context instead of replaying the whole conversation, and adapts its cadence — so the "learn from what just happened" loop that runs after your turns costs a fraction of what it used to. You keep the self-improvement, you stop paying full main-model price for it. (#49252 — @teknium1)
• •Compose your next prompt in your editor — `/prompt` — /prompt opens your $EDITOR so you can hand-write a long, multi-line prompt in real markdown instead of fighting a one-line input box. Draft a detailed spec, a structured question, or a big paste, save, and it's queued as your next message. Small thing, huge quality-of-life win for anyone who writes Hermes more than a sentence at a time. (#50509 — @teknium1)
• •Google Vertex AI — Gemini through your GCP service account, no static key — Vertex AI is now a first-class provider for Gemini models over Vertex's OpenAI-compatible endpoint. The reason a plain custom-provider setup always died mid-session is that Vertex has no static API key — every request needs a short-lived OAuth2 access token (~1h TTL) minted from a service-account JSON or Application Default Credentials. Hermes now mints and auto-refreshes those tokens for you, so if your org runs Gemini through Google Cloud, you point Hermes at your service account and it just works — no token-pasting, no mid-session expiry. (#56363 — @teknium1, @slawt)
• •Security round — This window hardened several surfaces: MCP-config persistence attack surface locked down, cron base_url overrides that could exfiltrate provider credentials blocked, a non-reusable sentinel for prefix secrets in file reads, Slack app-level (xapp-) token redaction, a browser cloud-metadata floor enforced on every backend, and an aiohttp CVE floor across the lazy messaging paths. Fewer ways for a prompt-injected or misconfigured session to leak a credential. (#50476, #56196, #54166, #56227, #52349, #56237 — @teknium1, @kshitijk4poor, @claudlos)
• •Presets as selectable virtual models — each named MoA preset appears as a model under provider moa; pick it in any model picker and Hermes routes through the ensemble (#46081, #53561, #53775 — @teknium1)
• •`/moa` is now one-shot sugar — runs a single prompt through the default preset and restores your model afterward; persistent switching goes through the model picker (#53548 — @teknium1)
• •Reference-model output shown as labelled blocks in CLI, TUI, and desktop — read each model's reasoning before the aggregator's synthesis (#53793, #53855 — @teknium1)

Hermes Agent v0.17.0 (v2026.6.19)

June 19, 2026

• •Hermes reaches iMessage — Photon Spectrum, no Mac relay required — There's now an iMessage platform plugin built on Photon's managed line pool. Run hermes photon login, authenticate with a device code, and Hermes can send and receive iMessage — no Mac sitting in a closet running a relay, no BlueBubbles bridge to babysit. It's positioned as the successor to BlueBubbles: free to start, nothing to self-host. If your friends and family live in the blue bubbles, Hermes lives there now too. (#32348, #42582, #44713 — @teknium1)
• •Raft — Hermes joins the Raft agent network as a gateway channel — A new bundled Raft platform adapter lets Hermes connect to Raft as an external agent through a wake-channel bridge. Set RAFT_PROFILE, run the bridge, and Raft can wake Hermes to handle messages — with a privacy-by-contract design where wake payloads carry only metadata (event IDs, timestamps), never message bodies. Another surface where Hermes can show up and do work. (#48210 — @xxchan, @teknium1)
• •A substantially more capable desktop app — v0.16.0 shipped the desktop app; v0.17.0 deepened it across dozens of PRs. Rebindable keyboard shortcuts, native OS notifications with per-type toggles, live subagent watch-windows that stream a delegated agent's activity into its own pane, a composer model selector with per-model presets, automatic RTL/bidi text direction, a resizable VS Code-themed terminal pane, per-thread composer drafts, and the ability to install any VS Code Marketplace theme directly into the app. The desktop is now a serious daily driver, not a preview. (#45866, #40660, #47060, #46959, #43292, #44596 — @OutThisLife, @teknium1)
• •Background / async subagents — delegate work and keep going — delegate_task(background=true) now dispatches a subagent that runs in the background and returns a handle immediately. You and the model keep working while it churns, and the full result re-enters the conversation as a new turn the moment it finishes. Kick off a long research dive or a multi-step build, then carry on with something else instead of sitting blocked waiting on it. (#40946, #46968 — @teknium1)
• •Edit images, not just generate them — image-to-image in `image_generate` — image_generate can now edit and transform a source image, not only create one from scratch. Pass an existing image and a prompt and it routes to the backend's edit endpoint (same tool, same pattern as video_generate), across every supported image provider. "Make this logo blue," "remove the background," "turn this sketch into a render" — all from the tool you already use. (#48705 — @teknium1)
• •Automation Blueprints — schedule things without learning cron — Pick an automation by name and Hermes asks you for what it needs — no cron syntax, no slot=value typing. One blueprint definition renders natively on every surface: a form in the dashboard, a slash command in the CLI/TUI/messenger, a conversation with the agent, an entry in the docs catalog. "Daily news briefing at 8am" becomes a thing you set up by answering questions, not by memorizing 0 8 * * *. (#41309 — @teknium1)
• •Cursor's Composer model, through your xAI Grok subscription — grok-composer-2.5-fast is now in the xAI OAuth model picker, with its context window reconciled to the full 200k. Composer is the fast coding model behind Cursor — and if you have an xAI Grok subscription, you can now point Hermes at it directly over OAuth, no separate API key. Your Grok plan, Hermes's agent loop, Composer's coding speed. (#47908, #6f89e17 — @teknium1)
• •Full profile builder in the dashboard — Build a complete Hermes profile from the browser — pick its model, choose its skills, attach its MCP servers — without hand-editing config.yaml. The dashboard also unified multi-profile management into one machine-wide view with a global profile switcher, so you manage every profile from a single place. (#39084, #44007 — @teknium1)
• •Skills Hub browser rehaul — The dashboard's Skills Hub got a ground-up rework: connected hubs, a Featured section, full skill previews before you install, and a security scan on each skill. Browsing and installing skills from the trusted taps (OpenAI, Anthropic, HuggingFace, NVIDIA) is now a real browsing experience, not a flat list. (#40384, #43398 — @teknium1)
• •The `memory` tool got a major upgrade — atomic batch operations — The memory tool gained an operations array that applies a batch of add/replace/remove edits atomically against the final character budget. The model can free up space and add new entries in a single call — even when an add alone would overflow the budget — collapsing what used to be a fragile multi-turn dance into one reliable operation. Memory updates are now faster and far less likely to fail mid-edit. (#48507 — @teknium1)
• •Secure dashboard login — The dashboard's authentication was hardened: every token-required endpoint now correctly returns 401 behind the OAuth gate, websocket auth uses the served dashboard token, and a warning fires when a public_url override is silently rejected. Exposing your dashboard to the network is safer by default. (#42578, #42578 — @benbarclay, @teknium1)
• •Official WhatsApp Business Cloud API adapter — Alongside the existing Baileys bridge, Hermes now speaks the official WhatsApp Business Cloud API — Meta's first-party, hosted, no-bridge-process path. Point it at your Business API credentials and Hermes talks WhatsApp through the supported channel, with no QR-scanning bridge process to keep alive. (#44331, #43921 — @jquesnelle, @teknium1)
• •Rich text for Telegram — Bot API 10.1 rich messages — Telegram replies now render as proper rich messages via Bot API 10.1: better formatting, cleaner long-message handling, native markup instead of flattened text. It's on by default with an opt-out, so your Telegram conversations look the way they should without any configuration. (#44829, #45584, #45953 — @teknium1)
• •Curator cost optimization — no aux-model spend on routine runs — The skill curator now prunes stale skills by default but no longer runs its LLM-powered consolidation pass unless you opt in (curator.consolidate: true or hermes curator run --consolidate). The deterministic inactivity sweep keeps running for free; the opinionated, aux-model-spending "build umbrella skills" fork is now off by default. Routine background curation costs you zero tokens. (#47840 — @teknium1)
• •Rebindable keyboard shortcuts panel; native OS notifications with per-type toggles; curated turn-completion cue + dismissable error banners (#40660, #45866, #42480, #47985 — @OutThisLife, @teknium1)

Hermes Agent v0.16.0 (2026.6.5) — The Surface Release

June 6, 2026

• •Hermes Desktop — a real native app, not a terminal wrapper — This is the headline. There's now a apps/desktop/ Electron application that installs like any other desktop app on macOS, Linux, and Windows, updates itself in place from inside the app, and gives you a polished GUI for everything Hermes does. You get a proper chat window with streaming, a session list you can archive and search, drag-and-drop files anywhere in the chat area, clipboard image paste, a Cmd+K command palette, and a model picker right in the status bar. If you've been telling friends "it's a CLI agent" and watching their eyes glaze over — now you can just send them an installer. None of this existed a week ago. (#20059, #35607, #37099, #37379, #38631 — @OutThisLife, @jquesnelle, @ethernet8023, @austinpickett, @benbarclay)
• •Run the desktop app against a remote Hermes — sign in with OAuth or username/password — The desktop app doesn't have to run Hermes locally. Point it at a remote Hermes gateway (your homelab, a hosted box, a teammate's server) and it connects over a secure WebSocket, authenticating with OAuth or a username/password login — no fiddling with --insecure flags or hand-copied session tokens. Each profile can target its own remote host, and you can run concurrent sessions across multiple profiles in one window with cross-profile @session links. The practical version: your laptop runs a thin GUI, the heavy agent runs wherever your API keys and compute live. (#37888, #38851, #39330, #39778 — @benbarclay, @OutThisLife, @teknium1)
• •The web dashboard is now a full admin panel — configure everything from the browser — The dashboard grew from "view your sessions" into a complete administration surface. There's a Channels page that sets up every gateway messaging platform (Telegram, Discord, Slack, etc.) from the browser, an admin panel for the MCP catalog with enable/disable toggles, credential management, webhook and hook creation, memory configuration, gateway controls, and a System page with check-before-update and one-click Debug Share. You no longer have to SSH in and edit config.yaml to wire up a new messaging channel or MCP server — it's all point-and-click now. (#36704, #36736, #37211, #38205, #38600 — @teknium1)
• •Hermes Desktop speaks Simplified Chinese — full 简体中文 in the chat GUI — The desktop app now ships a complete Simplified Chinese (简体中文) translation across every UI surface — the chat window itself, sidebar, settings, command center, cron, messaging, profiles, skills, agents, the lot. English stays the default; switch language in Appearance settings and the choice persists to your config (display.language). It's built on a proper typed i18n layer, so adding more languages from here is straightforward. (#38241 — @JimLiu)
• •Leaner default skill set — Hermes ships only what you actually need — The bundled skill set got a deliberate trim. Skills that were redundant or dead are gone (spotify — superseded by the native Spotify plugin's 7 tools; linear — superseded by hermes mcp install linear; kanban-codex-lane, debugging-hermes-tui-commands, a stale domain orphan, and several empty category markers). Heavier or niche skills moved from bundled to optional (the Baoyu creative set, dspy, subagent-driven-development, minecraft-modpack-server, pokemon-player, hermes-s6-container-supervision) — still one hermes skills install away, just not loaded by default. And a new environments: relevance gate keeps context-specific skills (kanban, docker/s6) out of the skills index for users who'll never use them, while still loading them on explicit request. The result: a smaller, sharper default skill list, less noise in the picker, and a lighter prompt. The curator can now prune unused built-in skills too (not just agent-created ones), with usage tracked for every skill. (#39028, #36701, #36228 — @teknium1)
• •NVIDIA/skills is now a built-in trusted skills tap — NVIDIA/skills joins OpenAI, Anthropic, and HuggingFace as a default trusted tap in the Skills Hub — discoverable, browsable, searchable, and auto-updating through the same pipeline. NVIDIA's verified skills for CUDA-X, AIQ, cuOpt and the rest of their product stack are one install away, with real category labels from the repo's skills.sh.json sidecar. (#34333 — @teknium1)
• •Quick Setup via Nous Portal — from install to first message in seconds — First-time setup was thinned down to two clear paths: Quick Setup (sign in with Nous Portal, get a model picker, start chatting immediately) or Full Setup (the detailed wizard for power users). hermes portal is now the human-readable alias that runs the full quick-setup Nous flow. The first-run menu explains the difference inline so newcomers aren't guessing. The goal: a brand-new user shouldn't need to read docs to send their first message. (#35723, #36227, #38449, #38465 — @teknium1, @kshitijk4poor)
• •Fuzzy model picker, everywhere — type a few letters, find your model — The model picker now does fuzzy search across the desktop app, web dashboard, TUI, and CLI. Type "v4fl" and deepseek-v4-flash surfaces; multi-endpoint providers are grouped under one row instead of cluttering the list with duplicates, and each row carries a description so you know what you're picking. The catalog refreshes hourly instead of daily, so new models show up the same day they launch. New this window: deepseek-v4-flash, MiniMax-M3 with 1M context, qwen3.7-plus. (#36928, #35227, #35756, #35659, #36214 — @kshitijk4poor, @teknium1)
• •`/undo [N]` — take back the last N turns — Said the wrong thing, or sent the agent down a bad path? /undo backs up N user turns, prefills your last message so you can edit and resend, and soft-deletes the turns in between. It works in the CLI, the TUI, and across messaging platforms (Telegram, Discord, etc.) with full parity. Closes a long-standing request (#21910). (#36229, #36699 — @teknium1)
• •Choose your default interface — `cli` or `tui` — You can now set whether hermes chat drops you into the classic CLI or the Ink TUI by default, with a --cli flag to override per-invocation. The TUI also got a single unified /model command and a Sessions overlay for switching between live sessions. Use the interface you actually like. (#37782, #37112 — @OutThisLife)
• •macOS desktop install + in-app self-update; rebuild-and-relaunch cleanly on macOS (#35607, #36198, #38296 — @OutThisLife)
• •macOS installer renamed to "Hermes" and turned into a launcher (#37516 — @OutThisLife)
• •Build desktop in its own desktop stage on macOS/Linux instead of silently skipping; content-hash build stamp, --build-only / --force-build flags (#36134 — @OutThisLife, #37597 — @ethernet8023)
• •Boot-failure recovery + live API-key validation; cancellable install; recover from corrupt cached Electron download (#35864, #37379, #39032 — @OutThisLife, @teknium1)
• •Windows: recover from corrupt Electron cache in bootstrap install; stop racing our own backend during in-app update (#39465, #39828 — @teknium1, @OutThisLife)

Hermes Agent v0.15.2 (2026.5.29.2)

May 29, 2026

• •Packaging): ship bundled plugin.yaml manifests in wheel and sdist (`827f7f07`)
• •@outsourc-e (1 commit)
• •@dparikh79 (1 commit)
• •@ousiaresearch (1 commit)
• •@libre-7 (1 commit)

Hermes Agent v0.15.1 (2026.5.29) — The Patch Release

May 29, 2026

• •Dashboard 401 reload loop fixed — In loopback mode the dashboard's identity probe (/api/auth/me) returns 401 by design, but v0.15.0's stale-token reload guard treated every 401 as a rotated session token and full-page-reloaded to pick up a fresh one. Every successful sibling call cleared the one-shot reload guard, so the page reload-looped forever (Firefox: "Navigated to /sessions" storm; Chrome: React re-render storm). Fix adds an allowUnauthorized opt-out to fetchJSON that skips only the loopback stale-token reload — 401 still throws so AuthWidget swallows it, gated-mode login_url redirects are unaffected. Closes #34206, #34202. (#30698 — @austinpickett)
• •Docker dashboard `--insecure` is now an explicit env opt-in, never derived from bind host — Previously the Docker entrypoint inferred --insecure when the dashboard bound to a non-loopback host. That conflated "I want LAN access" with "I want to disable the same-origin guard." The fix splits them: bind host is bind host, and disabling the dashboard's loopback auth requires an explicit HERMES_DASHBOARD_INSECURE=1. Existing setups that genuinely wanted insecure binding must now set the env var. (#34188, #34204 — @benbarclay)
• •MCP bare command resolution under Docker — MCP servers configured with bare commands (npx, npm, node) now resolve against /usr/local/bin so they actually launch inside the Docker image where those binaries live. v0.15.0 left these failing silently in containers when the agent's effective PATH didn't include the Node toolchain location. (#34186 — @benbarclay)
• •Skills page sidebar / source pills restored — A stale useMemo dependency in the new dashboard skills page collapsed the source pills and category sidebar to "All" only. Fixed; both surfaces now reflect the live catalog state. (#34194)
• •Kanban worker can be killed again — SIGTERM on a kanban worker was being absorbed by an intermediate process and the worker stayed running. Closes #28181. (#34045)
• •Full skills.sh catalog (858 → 19,932 entries) — The skills hub page was pulling a partial paginated catalog. The fetch now walks the sitemap, so all 19,932 skills.sh entries surface in the picker instead of just the first 858. (#34025)
• •`/api/auth/me` 401 no longer triggers reload loop in loopback mode — (#30698 — @austinpickett)
• •Skills page source pills + category sidebar restored — stale useMemo dep (#34194)
• •`--insecure` is now explicit opt-in via env var, not derived from bind host (#34188 — @benbarclay)
• •Dashboard test suite repaired to match the insecure-opt-in fix (#34204 — @benbarclay)
• •arm64 PR builds skip the GHA cache to avoid cache-thrash on cross-arch builders (#33704 — @BROCCOLO1D)
• •Bare `npx`/`npm`/`node` resolve against `/usr/local/bin` for Docker compatibility (#34186 — @benbarclay)
• •Worker SIGTERM actually terminates the process (#34045)
• •Workers receive images referenced in task bodies for vision-capable models (#34210)
• •`.md` files deliver again — media-delivery validation defaults to denylist-only instead of an overly-narrow allowlist (#34022)

Hermes Agent v0.15.0 (2026.5.28) — The Velocity Release

May 28, 2026

• •The Big Refactor — `run_agent.py` is no longer 16,000 lines — The file at the heart of Hermes — the agent conversation loop — has been reduced from 16,083 lines to 3,821 (-76%), with the extracted code redistributed across 14 cohesive modules under agent/. Behavior is unchanged: every extraction keeps a thin forwarder on AIAgent, every test patch path still works, every external caller is compatible. The reason you care: future Hermes development moves faster, plugin authors can finally grep the codebase, and the file that took 90 seconds to load in your editor opens in a blink. (#27248)
• •Kanban grew into a real multi-agent platform — 104 PRs end to end — Triage auto-decomposes one task into a tree of sub-tasks. hermes kanban swarm creates a full Swarm v1 graph in one command — root, parallel workers, gated verifier, gated synthesizer, shared blackboard. Tasks support per-task model overrides (cheap models for boilerplate, expensive ones for hard sub-tasks), board-level default workdirs, per-task worktree paths and branches, scheduled start times, configurable claim TTL, retry fingerprinting, stale-task detection, respawn guards, and a drag-to-delete trash zone. Workers report through /workers/active, /runs/{id}, and /inspect endpoints. (#27572, #28443, #28364, #28394, #28462, #28384, #28467, #28455, #28452, #28432, #28468, #28420)
• •Cold-start perf wave keeps going — another second saved, 47% fewer per-turn function calls — Three new optimization rounds: defer openai._base_client import (-240ms / -17MB on every CLI invocation), hot-path optimizations cut 47% of per-conversation function calls (399k → 213k for 31-turn chat), defer compression-feasibility check (-170 to -290ms on every agent construction), adaptive subprocess polling (-195ms per tool call, 1+ second per turn). Termux cold start drops from 2.9s to 0.8s. hermes --version cold drops 63% (701ms → 258ms), flipping the head-to-head benchmark against Codex CLI from 5/11 wins to 6/11. (#28864, #28866, #28957, #29006, #29419, #30121, #30609, #31968)
• •`session_search` rebuilt — no LLM, no cost, 4,500× faster — The old session_search was an aux-LLM-powered tool that cost ~$0.30/call and took ~30 seconds to summarize three sessions, sometimes confabulating when the right session wasn't even in the FTS5 hit list. The new shape is one tool with three modes (discovery, scroll, browse) inferred from which args are set — no mode parameter, no aux-LLM, no config knob, no companion skill. Discovery is ~20ms instead of ~90s; scroll is ~1ms. Searching your past sessions for context is now free and instant. (#27590)
• •Promptware defense — Brainworm-class attacks blocked at three chokepoints — Inspired by recent Brainworm / Promptware Kill Chain research (Origin HQ, arxiv 2601.09625), Hermes now defends the context window against prompt-injection attacks that try to hijack the agent via tool output, recalled memory, or stored skills. Single source of truth (tools/threat_patterns.py) with ~15 new Brainworm/C2 patterns; recalled memory is scanned at load time; tool results get delimiter markers so a malicious file or remote service can't impersonate Hermes' own system content. Paired with a new security-guidance plugin that pattern-matches dangerous code writes. (#32269, #33131, #9151)
• •Bitwarden Secrets Manager — one bootstrap token replaces every per-provider API key — Stop keeping plaintext API keys in ~/.hermes/.env. Install Bitwarden Secrets Manager (bws auto-installs lazily on first use), point Hermes at it with one bootstrap token (BWS_ACCESS_TOKEN), and every credential you need comes from Bitwarden at startup. Rotate a key in the Bitwarden web app and the rotation actually takes effect — Bitwarden defaults to source-of-truth so its values overwrite matching env vars on startup. Flip secrets.bitwarden.override_existing: false to invert. EU Cloud and self-hosted Bitwarden server URLs supported. Detected credentials are now labeled with their source so you can see at a glance which keys came from Bitwarden vs. the local env. (#30035, #31378, #30364)
• •ntfy as the 23rd messaging platform — push notifications without an account — ntfy is the self-hostable push-notification service with no signup, no API key, just a topic URL. Hermes now adapts to it as a platform plugin (zero edits to core), so your agent can send you push notifications from any cron job, kanban task completion, or chat send_message — to your phone, your watch, your desktop, your homelab. (salvages #30625 → originally #4043) (#30867)
• •Skill bundles — `/<name>` loads multiple skills at once — A skill bundle is a named group of skills that loads them all together with one slash command. Set up your "writing day" bundle (humanizer + ideation + obsidian + youtube-content) and /writing-day activates all four for the session. Skills Hub now has health checks, a freshness badge, and a watchdog cron. Three new optional skills land: code-wiki (Karpathy's LLM-Wiki, persistent indexed dev wiki), openhands (delegate to OpenHands for parallel coding agents), and web-pentest (OWASP-style web pentest recipes). (#28373, #32345, #32240, #32261, #32265)
• •TUI session orchestrator — multiple live sessions in one TUI window — The Ink TUI gained an active-session switcher overlay. List, switch between, refresh, and close multiple live process-local sessions without leaving the TUI; dispatch a new session with a session-scoped model picker. Plus a wave of TUI polish — mouse-tracking DEC mode presets, scrollback preservation across branches and termux, slash-dropdown fixes, x.com link rendering, and CJK / IME input rendering improvements. (salvages #27642) (#32980, #30084)
• •Two new image_gen providers — Krea 2 Medium + Large, FAL ported to plugin — Krea joins the image_gen lineup as a built-in plugin: Krea 2 Medium ($0.03) and Krea 2 Large ($0.06), auto-discovered, selectable via hermes tools → Image Generation → Krea. Available through both the native Krea plugin and the FAL.ai catalog. The FAL.ai backend got pulled out of the monolithic image-generation tool into plugins/image_gen/fal/, completing the four-way architectural parity already established by web, browser, and video_gen — new image providers are now one file, not a fork. (#33236, #30380, #33506)
• •Nous-approved MCP catalog with interactive picker — A curated catalog of Nous-vetted MCP servers, mirroring the optional-skills shape. Run hermes mcp and you get an interactive picker; install with one keystroke, credentials prompted at install time and written to ~/.hermes/.env. Ships with the n8n manifest first. Closes the discovery gap that left users hunting GitHub for trusted MCP servers. (#30870)
• •OpenHands orchestration skill — A new optional skill under optional-skills/autonomous-ai-agents/openhands/ lets the agent delegate coding tasks to the OpenHands CLI alongside claude-code, codex, and opencode. OpenHands is the model-agnostic member of that family — any LiteLLM-supported provider works (OpenAI, Anthropic, OpenRouter, your own), so you can route a sub-task to the cheapest model that can finish it. Drop-in worker for kanban swarms and /delegate flows. (closes #477) (#32261)
• •Deep xAI integration round — Web Search plugin, OAuth proxy upstream, May 15 retirement detection, natural TTS, security hardening — Six interlocking xAI improvements:
• •xAI Web Search lands as a plugins/web/xai/ provider, slots alongside Brave / Tavily / Exa / SearXNG / DDGS / Firecrawl — reuses your existing Grok OAuth or XAI_API_KEY credentials, no new env vars. (#29042)
• •`hermes proxy` gains an xAI upstream — your local OpenAI-compatible endpoint can now be backed by SuperGrok OAuth, no PKCE-refresh code to write in your client. (#28356)

Hermes Agent v0.14.0 (2026.5.16)

May 16, 2026

• •xAI Grok via SuperGrok OAuth — and grok-4.3 jumps to a 1M context window — If you pay for SuperGrok, you can now use Grok inside Hermes by signing in with your xAI account — no API key, no separate billing. The wire-through also bumps grok-4.3 to a 1M token context window, so you can drop whole codebases or research corpora into a single prompt. Includes proper handling for entitlement errors and an SSH-to-tunnel docs page for when you're SSH'd into a remote box and need to complete the OAuth flow. (#26534, #26664, #26644, #26592)
• •OpenAI-compatible local proxy for OAuth providers — Run hermes proxy and you get a http://localhost:port endpoint that speaks the OpenAI API but is backed by whichever OAuth provider you're signed into — Claude Pro, ChatGPT Pro, SuperGrok. Now any tool that expects an OpenAI-compatible endpoint (Codex CLI, Aider, Cline, Continue, your custom scripts) just works with your existing subscription, no API key required. One subscription, every tool. (#25969)
• •`x_search` — first-class X (Twitter) search tool — The agent can now search X directly without installing a skill or wiring up a custom integration. Search the timeline, find threads, surface specific posts — straight from the chat. Auth with either your X OAuth login or an API key, whichever you have. (#26763)
• •Microsoft Teams — end-to-end — Hermes can now read messages from Teams and post back. The full Microsoft Graph stack lands together: auth + client foundation, a webhook listener that receives Teams events, a pipeline plugin runtime, and outbound delivery. Wire up the bot once, then chat to your agent from any Teams channel, DM, or group. (salvages of #21408–#21411) (#21922, #21969, #22007, #22024)
• •Debloating wave — lighter installs, less you don't use — A clean pip install hermes-agent used to pull down everything: every messaging adapter SDK, every image-gen SDK, every voice/TTS provider, whether you used them or not. Now those heavy backends (Slack / Matrix / Feishu / DingTalk adapters, hindsight client, codex app-server, Pixverse / Camofox / image-gen SDKs, voice/TTS providers) install automatically the first time you actually use them. The [all] extras drop everything covered by lazy-deps, the installer falls back through tiers when a wheel doesn't fit your platform, and a supply-chain advisory checker scans every install for unsafe versions. Faster installs, smaller disk footprint, fewer transitive vulnerabilities. (#24220, #24515, #25014, #25038, #25766, #21818)
• •`pip install hermes-agent && hermes` — Hermes Agent is now a real PyPI package. No more cloning the repo or running shell installers — one pip command and you're running. The wheel ships with the Ink TUI bundle and the shell launcher, so the full experience comes out of the box. (salvage of #26350) (#26593, #26148)
• •Cross-session 1h Claude prompt cache — When you use Claude through Anthropic, OpenRouter, or Nous Portal, the prompt prefix (system prompt, skills, memory) now caches for an hour across sessions. Start a /new session and the first response comes back faster and cheaper because the cache is still warm from your last session. Background memory review hits the cache too, so it's not paying full price every turn. (#23828, #25434, #24778)
• •180x faster `browser_console` evaluations — When the agent uses the browser tool to inspect a page or run JavaScript, those calls now share one persistent connection to Chrome instead of spinning up a new DevTools session every time. The difference is huge: things that used to take a couple of seconds per call return in milliseconds. Real-world page interactions feel instant. (#23226)
• •Cold-start performance wave — ~19 seconds off `hermes` launch — Running hermes used to make you wait through a chunk of import overhead and network calls before you saw a prompt. Now the launch path is mostly deferred: heavy adapters only load when you use them, model catalogs come from disk cache first, doctor checks run in parallel, and chat -q skips the welcome banner entirely. The hermes tools All-Platforms screen alone dropped from 14 seconds to under 1.5 seconds. (#22138, #22120, #22681, #22790, #22808, #22831, #22859, #22904, #22766, #25341)
• •Two new messaging platforms — LINE + SimpleX Chat — LINE is huge in Japan, Korea, and Taiwan, and now Hermes runs natively on the LINE Messaging API. SimpleX Chat is the privacy-focused decentralized messenger with no user IDs — also wired up as a first-class platform. That brings Hermes to 22 messaging platforms total, so wherever you and your team chat, the agent can be there. (#23197, #26232)
• •`/handoff` actually transfers the session live — Switching models or personalities mid-conversation used to mean losing context or starting over. Now /handoff moves your active session — every message, every tool call, every piece of context — to the target model, persona, or profile, live, without dropping anything. Mid-debugging hand off from a fast model to a deep-reasoning one, or pass a session between profiles for different parts of a task. (#23395)
• •Native button UI for `clarify` on Telegram and Discord — When the agent uses the clarify tool to ask you a multiple-choice question, it now shows real platform-native buttons on Telegram and Discord instead of asking you to type back the option number. Tap the button, the agent gets your answer. Especially nice on mobile. (#24199, #25485)
• •Discord channel history backfill (default on) — When Hermes joins a Discord channel or thread for the first time, it now reads the recent message history so it knows what's been said before it responds. No more "what are we talking about?" — the agent has the context that's already on screen for everyone else. (#25984)
• •`vision_analyze` returns pixels to vision-capable models — When you point the agent at an image with vision_analyze and the active model can actually see (GPT-5, Claude, Gemini, Grok-vision), Hermes now passes the raw pixels straight to the model instead of converting them to a text description first. You get the model's actual visual reasoning instead of a degraded text-summary round-trip. (#22955)
• •Per-turn file-mutation verifier footer — After every turn that wrote or edited files, the agent now gets a short footer summarizing exactly what changed on disk — the file paths, the line counts, the actual delta. That means the agent catches its own mistakes when a write didn't land or got silently overwritten, instead of confidently telling you "I added the function" when the file wasn't actually saved. (#24498)

Hermes Agent v0.13.0 (2026.5.7) — The Tenacity Release

May 7, 2026

• •Multi-agent Kanban — delegate to an AI team that actually finishes — Spin up a durable board, drop tasks on it, and let multiple Hermes workers pick them up, hand off, and close them out. Heartbeats, reclaim, zombie detection, retry budgets, and a hallucination gate keep the team honest. One install, many kanbans. (#17805, #19653, #20232, #20332, #21330, #21183, #21214)
• •`/goal` — the agent doesn't forget what you asked it to do — Lock the agent onto a target and it stays on task across turns. The Ralph loop as a first-class primitive. (#18262, #18275, #21287)
• •Show it a video — new video_analyze tool for native video understanding on Gemini and compatible multimodal models. (@alt-glitch) (#19301)
• •Clone a voice — xAI Custom Voices lands as a TTS provider with voice cloning support. (@alt-glitch) (#18776)
• •Hermes speaks your language — static gateway + CLI messages translate to 7 locales: Chinese, Japanese, German, Spanish, French, Ukrainian, and Turkish. Docs site gains a Chinese (zh-Hans) locale. (#20231, #20329, #20467, #20474, #20430, #20431)
• •Google Chat — the 20th messaging platform — plus a generic platform-plugin hooks surface so third-party adapters drop in without touching core (IRC and Teams migrated). (#21306, #21331)
• •Sessions survive restarts — gateway bounces mid-agent, /update restarts, source-file reloads — conversations auto-resume when the gateway comes back. (#21192)
• •Security wave — 8 P0 closures — redaction ON by default, Discord role-allowlists guild-scoped (CVSS 8.1 cross-guild DM bypass closed), WhatsApp rejects strangers by default, TOCTOU windows closed across auth.json and MCP OAuth, browser enforces cloud-metadata SSRF floor, cron prompt-injection scans assembled skill content, hermes debug share redacts at upload. (#21193, #21241, #21291, #21176, #21194, #21228, #21350, #19318)
• •Checkpoints v2 — state persistence rewritten. Real pruning, disk guardrails, no more orphan shadow repos. (#20709)
• •The agent lints its own writes — post-write delta lint on write_file + patch. Python, JSON, YAML, TOML. Syntax errors surface immediately instead of shipping downstream. (#20191)
• •`no_agent` cron mode — script-only watchdog — cron jobs can now skip the agent entirely and just run a script. Empty stdout is silent, non-empty gets delivered verbatim. (#19709)
• •Platform allowlists everywhere — allowed_channels / allowed_chats / allowed_rooms config across Slack, Telegram, Mattermost, Matrix, and DingTalk. (#21251)
• •Providers are now plugins — ProviderProfile ABC + plugins/model-providers/. Drop in third-party providers without touching core. (#20324)
• •API server — long-term memory per session — X-Hermes-Session-Key header gives memory providers a stable session identifier. (#20199)
• •MCP levels up — SSE transport with OAuth forwarding, stale-pipe retries, image results surface as MEDIA tags instead of getting dropped, keepalive on long-lived lifecycle waits. (#21227, #21323, #21289, #21328, #20209)

Hermes Agent v0.12.0 (2026.4.30)

April 30, 2026

• •Autonomous Curator — hermes curator runs as a background agent on the gateway's cron ticker (7-day cycle default). It grades your skill library, consolidates related skills, prunes dead ones, and writes per-run reports to logs/curator/run.json + REPORT.md. Archived skills are classified consolidated-vs-pruned via model + heuristic. Defense-in-depth gates protect bundled/hub skills from mutation. Unified under auxiliary.curator — pick the curator's model in hermes model, manage it from the dashboard. hermes curator status ranks skills by usage (most-used / least-used). (#17277, #17307, #17941, #17868, #18033)
• •Self-improvement loop — substantially upgraded — The background review fork (the core of Hermes' self-improvement: after each turn it decides what memories/skills to save or update) is now class-first (rubric-based rather than free-form), active-update biased (prefers the skill the agent just loaded), handles references//templates/ sub-files, and properly inherits the parent's live runtime (provider, model, credentials actually propagate). Restricted to memory + skills toolsets so it can't sprawl. Memory providers shut down cleanly. Prior-turn tool messages excluded from the summary so the fork sees a clean context. (#16026, #17213, #16099, #16569, #16204, #15057)
• •Skill integrations — major expansion — ComfyUI v5 with official CLI + REST + hardware-gated local install, moved from optional to built-in by default (#17610, #17631, #17734). TouchDesigner-MCP bundled by default, expanded with GLSL, post-FX, audio, geometry, and 9 new reference docs (#16753, #16624, #16768 — @kshitijk4poor + @SHL0MS). Humanizer skill ports a text-cleaner that strips AI-isms (#16787). claude-design HTML artifact skill + design-md (Google DESIGN.md spec) + airtable salvage + skill_manage edits in external_dirs + direct-URL skill install + /reload-skills slash command. (#16358, #14876, #16291, #17512, #16323, #17744)
• •LM Studio — first-class provider — upgraded from a custom-endpoint alias to a full-blown native provider: dedicated auth, hermes doctor checks, reasoning transport, live /models listing. (Salvage of @kshitijk4poor's #17061.) (#17102)
• •Four more new inference providers — GMI Cloud (first-class, salvage of #11955 — @isaachuangGMICLOUD), Azure AI Foundry with auto-detection, MiniMax OAuth with PKCE browser flow (salvage #15203), Tencent Tokenhub (salvage of #16860). (#16663, #15845, #17524, #16960)
• •Pluggable gateway platforms + Microsoft Teams — the gateway is now a plugin host. Drop-in messaging adapters live outside the core, and Microsoft Teams is the first plugin-shipped platform. (Salvage of #17664.) (#17751, #17828)
• •Tencent 元宝 (Yuanbao) — 18th messaging platform — native gateway adapter with text + media delivery. (#16298, #17424)
• •Spotify — native tools + bundled skill + wizard — 7 tools (play, search, queue, playlists, devices) behind PKCE OAuth, interactive setup wizard, bundled skill, surfacing in hermes tools, cron usage documented. (#15121, #15130, #15154, #15180)
• •Google Meet plugin — join calls, transcribe, speak, follow up. Realtime OpenAI transport + Node bot server, full pipeline bundled as a plugin. (#16364)
• •`hermes -z` one-shot mode + `hermes update --check` — non-interactive hermes -z <prompt> with --model/--provider/HERMES_INFERENCE_MODEL. hermes update --check preflight. Opt-in pre-update HERMES_HOME backup. (#15702, #15704, #15841, #16539, #16566)
• •Models dashboard tab + in-browser model config — rich per-model analytics, switch main + auxiliary models from the dashboard. (#17745, #17802)
• •Remote model catalog manifest — OpenRouter + Nous Portal model catalogs are now pulled from a remote manifest so new models show up without a release. (#16033)
• •Native multimodal image routing — images now route based on the model's actual vision capability rather than provider defaults. (#16506)
• •Gateway media parity — native multi-image sending across Telegram, Discord, Slack, Mattermost, Email, and Signal; centralized audio routing with FLAC support + Telegram document fallback. (#17909, #17833)
• •TUI catches up to (and past) the classic CLI — LaTeX rendering (@austinpickett), /reload .env hot-reload, pluggable busy-indicator styles (@OutThisLife, #13610), opt-in auto-resume of last session, expanded light-terminal auto-detection, session delete from /resume picker with d, modified mouse-wheel line scroll, and a /mouse toggle that kills ConPTY's phantom mouse injection (@kevin-ho). (#17175, #17286, #17150, #17130, #17113, #17668, #17669, #15488)