Skip to content
Agents tracked: 501 Releases (7d): 777 Agent status: no reported issuesSince last limit reset: Claude 10d · Codex 8d For AI agents: Markdown Updated Nov 21, 2026

Hermes Agent changelog: what's new each month

Every stable Hermes Agent release summarised by month: the highlights, new features, improvements, fixes and anything you need to act on. 9 months covered; the current month updates daily.

November 2026 · so far

6 releases: 0.22.6 → 0.23.3

November brought major architectural changes to Hermes Agent, unifying how conversations are managed across different interfaces and adding powerful new capabilities for collaboration and automation. The month focused on making the agent more accessible and controllable across browsers, terminals, and remote systems.

Highlights

  • Desktop workspace now accessible in any browser on your network through the new Webapp.
  • Single gateway now owns every conversation, eliminating message duplication across CLI, Desktop, editors and bots.
  • Images now display directly in terminal interfaces instead of showing file paths.
  • Rewind conversations to edit or fork from earlier messages in TUI and CLI.
  • Bot Mode rooms can now run themselves with automatic routing.

New

  • Webapp serves the full Desktop workspace from your gateway in any browser.
  • Rewind feature lets you pick an earlier message, edit it, and resend or fork a new session.
  • Desktop keeps all gateways connected at once for instant switching between local and remote machines.
  • Bot Mode rooms support automatic routing and lead bot selection.
  • Plugin auto-update option keeps catalog and git plugins current automatically.
  • Price tags and spend limits for paid MCP tools.
  • Pictures display inline in kitty and Ghostty terminals.
  • hermes plugins update command updates all plugins at once.

Improved

  • Desktop can now manage multiple gateway connections simultaneously.
  • Group Chats can operate autonomously with routing capabilities.
  • Plugin management streamlined with centralized update controls.
  • Terminal image rendering shows real pixels instead of file paths.
  • Webapp login experience improved for Hermes Cloud agents.
  • Chat persistence improved during Desktop startup and reconnection.
  • Reply streaming now settles properly when alongside todo or memory calls.
  • Desktop no longer loses typed content while resuming chats.

Fixed

  • Desktop chats created by August builds with bare custom providers can send again.
  • Duplicate "Recovered reply" messages no longer appear on Telegram after mid-stream crashes.
  • Desktop no longer scrambles or loses text typed during startup and resume.
  • Webapp correctly instructs unsigned users to sign into Nous Portal.
  • CLI no longer leaves git running after startup update check.
  • Security fix for Bot Desktop running in docker or ssh terminal backends.

October 2026

8 releases: 0.21.6 → 0.22.5

October brought major usability improvements and security fixes to Hermes Agent, including a redesigned setup experience, connectors for hosted apps, and support for running local models without API keys. The month also established stable release channels with SemVer versioning and grew the plugin catalog to 600 reviewed entries.

Highlights

  • First-launch setup chat replaces configuration screens for new users on desktop.
  • Connectors let Hermes act directly in apps like Linear and Notion through the Nous tool gateway.
  • Local models now run entirely on your machine with no account or API key required.
  • Browser agent can log into websites using passwords from 1Password, Bitwarden, or Hermes's encrypted vault.
  • Plugin catalog reached 600 reviewed and pinned plugins installable from desktop or CLI.

New

  • Setup chat on first launch of the desktop app instead of settings screen.
  • App connectors through Nous tool gateway for services like Linear and Notion.
  • Local model support for running open models entirely on your machine.
  • Password manager integration for browser-based logins (1Password, Bitwarden, Hermes vault).
  • Stable SemVer release channels with tested builds across all platforms before publication.
  • Signed desktop packages for Windows and macOS with self-contained Python and Node runtimes.
  • Update channel selection in desktop app to choose between stable releases or every commit.
  • Support for Claude 5.5, GPT-6, and Gemini 4 model variants on Nous Portal and OpenRouter.

Improved

  • Reasoning now survives chat resume and replays signed thinking on Bedrock Converse models.
  • Updates work reliably behind regional firewalls that previously blocked release record fetches.
  • Local model turns on laptops survive sleep by streaming with progress indicators and adjusted stall timers.
  • Plugin installation skips consent step for plugins that declare no dependencies.
  • Desktop app updates itself on Microsoft Store without manual intervention.
  • Gateway correctly restarts after system reboot on Linux by avoiding PID confusion.
  • Curated model list appears at the top of the model picker for easier selection.
  • WhatsApp pairing can be safely re-established on the dashboard after logout.

Fixed

  • Spoofed X-Forwarded-For headers could bypass password-login rate limits; now fixed.
  • Unauthenticated login requests wrote unbounded values to auth audit log; now capped.
  • Public /auth/ routes lacked request-body size limits; now enforced.
  • Native sign-in could send login codes to non-loopback redirects enabling session takeover; now restricted.
  • Automatic git calls could execute untrusted filter programs from repositories; now prevented.
  • Suggestion pills above composer appeared unnecessarily and have been removed.

September 2026

5 releases: 2026.9.7 → 2026.9.24

September brought major stability improvements to Hermes Agent, focusing on database concurrency issues, state corruption handling, and multi-platform message rendering. The month saw fixes to critical wedging problems and graceful degradation for edge cases that previously crashed sessions.

Highlights

  • Fixed database locking issues that caused 4–20 second stalls when opening state.db.
  • Eliminated sticky database errors from WAL generation and concurrent writes that left the store permanently wedged.
  • Made full-text-search failures graceful; corrupted search indexes no longer fail entire conversations.
  • Prevented single malformed database rows from crashing sessions list, export, or insights views.
  • Stopped profile database cross-contamination from Desktop launch and session search operations.

New

  • Matrix now renders LaTeX.
  • Signal renders markdown tables.
  • WhatsApp replies to messages (platform support).
  • Telegram bots enforce @mention requirement when botsrequiremention is enabled.

Improved

  • Remote dashboard sessions no longer expire during refresh bursts.
  • Long-lived processes no longer leak duplicate state.db writer handles in gateway, dashboard, Desktop backend, and ACP.
  • Opening state.db skips acquiring write lock when no writes are needed.
  • Queued-lane finals refused by the platform are now recorded and redelivered.
  • Stalled WebSocket sends are handled more robustly.
  • Full-text-search errors now degrade gracefully instead of failing the entire conversation.

Fixed

  • OpenZFS deleted dentries and close() racing appendmessage no longer produce sticky DeletedWalGenerationError.
  • Bare display key in config.yaml no longer crashes gateway turns.
  • TEXT timestamp or 1e30 epoch values no longer crash sessions listing.
  • Malformed marker JSON no longer crashes jsonextract operations.
  • More than 999 IDs in a listing no longer causes crashes.
  • Sessions no longer bind to or read another profile's database under HERMESHOME override race conditions.

August 2026

8 releases: 2026.8.3 → 2026.8.31

August brought major voice and collaboration features to Hermes, including streaming conversational voice with hands-free wake words, Agent-to-Agent communication, and Bot Mode for multi-agent coordination. The month also expanded platform reach and added powerful new orchestration and CLI tools.

Highlights

  • Streaming voice mode with barge-in lets you interrupt Hermes mid-reply, and custom wake words enable hands-free control on-device.
  • Voice messages now work across WhatsApp, Feishu, DingTalk, LINE, QQ, Photon, and Weixin with platform-aware replies.
  • Hermes agents can now discover, talk to, and coordinate with other Agent-to-Agent compatible agents via A2A v1.0.
  • Bot Mode makes your agent profiles into a society with persistent memory, cron scheduling, and bot-to-bot messaging.
  • Research output now includes grounded citations with fact-checking so every claim points to a verifiable source.

New

  • Wake words and hands-free control with on-device detection.
  • Outbound webhooks push signed lifecycle events to your systems instead of requiring polling.
  • Grounded citations skill backs research claims with verifiable sources.
  • Desktop app renders artifacts as versioned cards with sandboxed live preview.
  • Agent-to-Agent v1.0 protocol enables agent discovery and inter-agent communication.
  • Bot Mode with deterministic avatar faces and bot-to-bot DMs between agents.
  • Live orchestration for subagents: list running children, steer them mid-flight, or stop early.
  • In-app browser now directly navigated and clicked by agents, with page pop-out support.

Improved

  • Voice input works on every major messaging platform with transcription and auto-TTS replies.
  • Desktop app became a plugin platform with artifact viewer and quick entry.
  • Cron jobs now load and update persistent memory, carrying output between runs.
  • MCP servers and catalog merged into one coherent page with health checks and re-auth nudges.
  • CLI added fuzzy command palette (Ctrl+P), /model picker with filtering, and /status showing reasoning mode and context usage.
  • Six new model providers added: Meta Model API, CommandCode, Tencent TokenPlan, Nebius Token Factory, Ramp Router, and Actual.
  • CLI commands run instantly with !command without spending a model turn.
  • /init generates or updates AGENTS.md by scanning your project; /diff shows staged, all, or session changes.

July 2026

5 releases: 2026.7.1 → 2026.7.30

July brought major performance improvements and new ways to work with Hermes, including faster cold starts, smarter approvals, and deeper integration with your development workflow. The agent now learns continuously and can manage complex multi-task delegations without blocking your work.

Highlights

  • Mixture-of-Agents is now a selectable model option alongside others, with full visibility into each reference model's reasoning.
  • Cold-start time dropped from 4.3 seconds to 0.9 seconds, with a 20+ performance optimization wave across the desktop app.
  • Smart approvals now use independent LLM review instead of asking you to manually approve every flagged command.
  • Background task delegation lets you dispatch multiple subagents without blocking your chat, with live transcript monitoring.
  • Desktop app now has first-class Projects with sidebar codebase management, coding rail, review pane, and git worktree handling.

New

  • Mixture-of-Agents presets show as selectable models under a dedicated moa provider.
  • /learn command turns anything (directory, URL, description) into a reusable skill.
  • /journey command displays a timeline of memories and skills Hermes has learned about you.
  • Background fan-out delegation lets multiple subagents run in parallel with live transcript files you can monitor.
  • Password manager integration with Bitwarden and 1Password support via pluggable SecretSource interface.
  • Profile-based message routing allows one gateway to serve multiple profiles across different channels or guilds.
  • Delivery-obligation ledger ensures responses cannot be lost between generation and platform confirmation.
  • Agent now verifies its own work on coding projects by running them instead of claiming completion.

Improved

  • Hermes renders each reference model's full output as a labelled block in MoA ensembles so you can see what each model thought.
  • Desktop app markdown splitter reduced CPU cost by 14× for long replies.
  • Review pane no longer freezes on giant diffs thanks to virtualization.
  • Gateway can now go dormant when idle and shut down cleanly without dropping in-flight requests.
  • Subscription and top-up management moved from billing website into /subscription and /topup terminal commands.
  • Smart approvals now use independent LLM assessment of flagged commands instead of requiring manual approval.
  • Subagent delegation now returns live transcript files you can tail in real time.

Fixed

  • WhatsApp bridge dependency now installs from published npm release instead of a pinned git commit.
  • Agent no longer loses answers if the gateway dies between generating a response and confirming delivery.
  • Cold-start initialization time reduced from 4.3 seconds to 0.9 seconds.

June 2026

2 releases: 2026.6.5 → 2026.6.19

June brought Hermes Desktop as a native application, expanded platform connectivity, and substantial improvements to administration and automation capabilities. The month focused on accessibility for new users and deeper integrations with external services.

Highlights

  • Hermes Desktop launched as a native app for macOS and Linux, with support for local or remote Hermes instances via OAuth or username/password.
  • Hermes now connects to iMessage through Photon Spectrum and joins the Raft agent network as a gateway channel.
  • The web dashboard evolved into a full admin panel for configuring everything from channels to the complete profile builder.
  • Fuzzy model search is now available across desktop, web, TUI, and CLI for faster model selection.
  • Background subagents let you delegate work and continue operating without waiting for results.

New

  • Hermes Desktop native application for macOS and Linux with remote gateway support.
  • iMessage platform plugin via Photon Spectrum without requiring a Mac relay.
  • Raft agent network integration as an external gateway through a wake-channel bridge.
  • Background async subagents with immediate return handles for delegated tasks.
  • Image-to-image editing and transformation in imagegenerate beyond text-to-image creation.
  • Automation Blueprints for scheduling tasks without learning cron syntax.
  • Fuzzy model picker across all interfaces for faster model discovery.
  • Simplified Chinese (简体中文) translation in the desktop application.

Improved

  • Desktop app v0.17.0 added rebindable keyboard shortcuts, native OS notifications with per-type toggles, and dozens of capability enhancements.
  • Web dashboard expanded from session viewing to complete administration surface with channels configuration.
  • Quick Setup process simplified to two clear paths: Quick Setup via Nous Portal or standard setup.
  • NVIDIA/skills added as a built-in trusted skills tap alongside OpenAI, Anthropic, and HuggingFace.
  • Default skill set streamlined by removing redundant or obsolete skills.
  • Profile builder moved to the browser dashboard, eliminating need to hand-edit config.yaml.
  • Grok Composer model added through xAI OAuth with full 200k context window reconciliation.
  • Model picker now provides fuzzy search functionality across desktop, web, TUI, and CLI.

Fixed

  • Desktop app stability and UI responsiveness improved across v0.16.0 to v0.17.0.
  • iMessage connectivity resolved through managed line pool architecture.

May 2026

5 releases: 2026.5.7 → 2026.5.29.2

May brought major platform expansion, multi-agent coordination, and a significant performance overhaul. Hermes added 23 messaging platforms, introduced durable task boards for agent teams, and reduced core complexity by 76%.

Action needed
  • Discord role-allowlists are now guild-scoped instead of global.
  • Message redaction is on by default.
  • WhatsApp now rejects messages from strangers by default.
  • Lighter default pip install removes optional dependencies—install extras explicitly if needed.

Highlights

  • Multi-agent Kanban boards let you delegate work to a team of Hermes agents that coordinate, hand off, and complete tasks automatically.
  • The /goal command keeps agents focused on your original request across multiple turns without losing context.
  • Hermes Agent is now a standard pip install—no repo cloning required.
  • Added native video understanding, X (Twitter) search, Microsoft Teams integration, and 20 other messaging platforms.
  • Session search became 4,500× faster with zero LLM cost by removing the auxiliary model dependency.

New

  • Multi-agent Kanban boards with task decomposition and automatic worker coordination.
  • Video analysis tool for native video understanding on Gemini and compatible multimodal models.
  • xAI Custom Voices TTS provider with voice cloning support.
  • xAI Grok via SuperGrok OAuth with 1M context window.
  • Microsoft Teams read and post capability with full Microsoft Graph integration.
  • ntfy messaging platform adapter for self-hosted push notifications.
  • Skill bundles to load multiple skills together with one command.
  • OpenAI-compatible local proxy for OAuth providers (Claude Pro, Grok, others).

Improved

  • Localization expanded to 7 languages: Chinese, Japanese, German, Spanish, French, Ukrainian, and Turkish.
  • Sessions now auto-resume when the gateway restarts or source files reload.
  • Agent browser interactions 180× faster by sharing a persistent Chrome connection.
  • Claude prompt cache now persists for 1 hour across sessions, reducing redundant processing.
  • Cold-start performance improved by another second with fewer per-turn function calls (47% reduction).
  • Bitwarden Secrets Manager integration eliminates plaintext API keys from config files.
  • Generic platform-plugin hook surface allows third-party messaging adapters without core changes.
  • Skills hub catalog expanded from 858 to 19,932 entries by walking the full sitemap.

Fixed

  • Dashboard 401 reload loop in loopback mode eliminated.
  • Docker dashboard --insecure now requires explicit environment opt-in instead of inferring from bind host.
  • MCP bare command resolution fixed to properly resolve commands inside Docker images.
  • Kanban worker SIGTERM handling corrected so workers terminate as expected.
  • Skills page source pills and category sidebar now display correctly after UI update.
  • Eight P0 security closures including cross-guild DM bypass and TOCTOU windows.

April 2026

6 releases: 2026.4.3 → 2026.4.30

April brought major extensibility improvements to Hermes, with pluggable memory and context systems, expanded platform support across messaging and inference providers, and a complete rewrite of the interactive CLI. The agent also gained real-time capabilities like live model switching and streaming tool progress.

Highlights

  • Pluggable Memory Provider Interface and Context Engine let you swap in custom backends for memory and context management.
  • New Ink-based TUI completely rewrites the interactive CLI with live streaming, sticky composer, and a Python JSON-RPC backend.
  • 17th and 18th messaging platforms added: QQBot and Tencent Yuanbao, plus full iMessage support via BlueBubbles and WeChat integration.
  • Live model switching mid-session with the /model command, plus GPT-5.5 availability through ChatGPT Codex OAuth and eight new inference providers.
  • Autonomous Curator background agent now automatically grades, consolidates, and prunes your skill library on a 7-day cycle.

New

  • Pluggable Memory Provider Interface for extensible memory backends with automatic provider registration.
  • Camofox Anti-Detection Browser Backend for stealth browsing with persistent sessions and VNC debugging.
  • Background Process Auto-Notifications to alert the agent when long-running tasks finish.
  • Local Web Dashboard for managing settings, monitoring sessions, browsing skills, and controlling the gateway.
  • Fast Mode (/fast) for priority processing on OpenAI and Anthropic models with significantly lower latency.
  • Termux / Android Support to run Hermes natively on Android with mobile-optimized TUI and voice backend.
  • Nous Tool Gateway providing automatic access to web search, image generation, text-to-speech, and browser automation for paid subscribers.
  • /steer command for mid-run agent nudges that inject guidance after the next tool call without interrupting the turn.

Improved

  • Same-Provider Credential Pools with automatic rotation and thread-safe load distribution across multiple API keys.
  • Inline Diff Previews in the tool activity feed show visual confirmation of file writes and patches before the agent proceeds.
  • API Server Session Continuity with real-time tool progress streaming and persistent session support via X-Hermes-Session-Id headers.
  • Gateway Hardening with major stability improvements across race conditions, media delivery, flood control, stuck sessions, and approval routing.
  • Inactivity-Based Agent Timeouts now track actual tool activity instead of wall-clock time, preventing long-running active tasks from being killed.
  • Approval Buttons on Slack & Telegram using native platform buttons instead of typed commands, with thread context preservation.
  • Self-Optimized GPT/Codex Tool-Use Guidance through automated behavioral benchmarking fixing five failure modes in tool calling.
  • Plugin surface expanded with slash command registration, direct tool dispatch, tool execution veto hooks, and tool rewriting capabilities.

Fixed

  • Secret Exfiltration Blocking now scans browser URLs and LLM responses to block secrets exposed via URL encoding, base64, or prompt injection.
  • Gateway race conditions, photo media delivery issues, flood control, stuck sessions, and approval routing problems resolved in major stability pass.
  • MCP OAuth 2.1 PKCE implementation with automatic malware scanning of MCP extension packages via OSV vulnerability database.
  • Live model discovery wired into model picker so new OpenAI releases automatically appear without manual updates.
  • Crashed sessions in iMessage via BlueBubbles integration now handled with resilience improvements.
  • Self-improvement loop substantially upgraded to better decide what memories and skills to save or update after each turn.

March 2026

5 releases: 2026.3.12 → 2026.3.30

March brought major infrastructure overhauls and platform expansion to Hermes Agent. The release adds multi-platform messaging, streaming responses, plugin extensibility, and support for dozens of new inference providers and messaging adapters.

Highlights

  • Multi-platform messaging gateway now supports 10+ platforms including Telegram, Discord, Slack, WhatsApp, Signal, Email, and emerging platforms like Feishu and WeCom.
  • Real-time token-by-token streaming across CLI and all messaging platforms delivers responses as they're generated instead of in blocks.
  • First-class plugin architecture lets you drop Python files into ~/.hermes/plugins/ to extend Hermes with custom tools and commands without forking.
  • Native Anthropic provider support with Claude Code integration and prompt caching, plus access to 400+ models through Nous Portal and Hugging Face.
  • Multi-profile support runs multiple isolated Hermes instances from one installation, each with separate config, memory, and gateway service.

New

  • MCP (Model Context Protocol) client with stdio and HTTP transports for resource and prompt discovery.
  • Skills ecosystem with 70+ bundled skills across 15+ categories and a Skills Hub for community discovery.
  • Voice mode with push-to-talk in CLI, voice notes in messaging platforms, and local Whisper transcription.
  • Filesystem checkpoints and rollback system that snapshots before destructive operations.
  • OpenAI-compatible API server exposing Hermes as a /v1/chat/completions endpoint with REST API for cron jobs.
  • MCP server mode exposing Hermes conversations to any MCP-compatible client like Claude Desktop or Cursor.
  • Official Docker container for running Hermes in both CLI and gateway modes.
  • Ordered fallback provider chain with automatic failover between multiple inference providers.

Improved

  • Unified streaming infrastructure delivers real-time responses in CLI and all gateway platforms.
  • Concurrent tool execution runs multiple independent tool calls in parallel, reducing latency.
  • Smart approvals system learns which commands are safe and remembers your preferences.
  • Centralized provider router replaces scattered provider logic with unified callllm() API.
  • Context compression overhauled with structured summaries, iterative updates, and token-budget protection.
  • Honcho memory integration with async writes, configurable recall modes, and multi-user isolation.
  • CLI visual customization with data-driven skin/theme engine featuring 7 built-in skins and custom YAML support.
  • Gateway prompt caching preserves Anthropic prompt cache across turns for cost reduction on long conversations.

Fixed

  • OpenAI model reliability improved with GPTTOOLUSEGUIDANCE to prevent models from describing actions instead of making tool calls.
  • Extensive linebreak and concatenation fixes for streaming mode display.
  • Stale budget warnings automatically stripped from OpenAI responses.
  • Supply chain hardening removed compromised litellm dependency and pinned all dependency versions.
  • Git worktree isolation enables safe parallel work on the same repository in isolated agent sessions.
  • PII redaction automatically scrubs personally identifiable information before sending context to LLM providers.

Summaries are written automatically from the official release notes (full changelog ↗); check the original notes before relying on a detail. Hermes Agent: pricing, features and alternatives · All changelogs