Hermes Agent is an open-source AI agent framework developed by Nous Research — the team behind the Hermes family of fine-tuned LLMs that have dominated OpenRouter leaderboards. Unlike most AI assistants that reset between sessions, Hermes was built around a single idea: an agent that grows with you. It learns from experience by creating reusable skills, maintains persistent memory across conversations, and runs everywhere you work — terminal, desktop app, Telegram, Discord, Slack, WhatsApp, Signal, and more.
Since its initial release in March 2026, Hermes Agent has grown to 208k+ GitHub stars, 37.7k forks, and over 14,000 commits from 370+ community contributors. The project is MIT-licensed, provider-agnostic (works with OpenRouter, Anthropic, OpenAI, Google Gemini, DeepSeek, xAI, local models, and 15+ others), and runs on Linux, macOS, Windows, and even Android via Termux.
What Makes Hermes Different
Hermes occupies the same category as Claude Code, OpenAI Codex, and OpenClaw — autonomous agents that use tool calling to interact with your system. What sets it apart is its self-improvement loop:
- Self-improving through skills — When Hermes solves a complex problem or discovers a workflow, it saves that knowledge as a reusable skill document. Skills accumulate over time, making the agent better at your specific tasks.
- Persistent memory across sessions — Hermes remembers who you are, your preferences, environment details, and lessons learned. Pluggable backends include built-in storage, Honcho dialectic user modeling, and Mem0.
- Multi-platform gateway — The same agent runs on Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Email, SMS, iMessage, and 10+ other platforms with full tool access — not just chat.
- Provider-agnostic — Swap models and providers mid-workflow without changing anything else. Credential pools rotate across multiple API keys automatically.
- Runs anywhere — Six terminal backends (local, Docker, SSH, Singularity, Modal, Daytona) let your agent live on a $5 VPS, a GPU cluster, or serverless infrastructure that costs nearly nothing when idle.
Feature Timeline
Here’s a chronological look at the major features introduced across each release:
| Release | Key Features Introduced |
|---|---|
| v0.2.0 — Mar 12, 2026 | Initial public release: CLI agent, basic tool calling, terminal/file/web tools |
| v0.3.0 — Mar 17, 2026 | Messaging gateway (Telegram, Discord), session persistence, skill system |
| v0.4.0 — Mar 23, 2026 | WhatsApp and Signal support, context compression, MCP server integration |
| v0.5.0 — Mar 28, 2026 | Memory system, user profiles, delegate_task for subagent spawning |
| v0.6.0 — Mar 30, 2026 | Cron scheduling, webhook subscriptions, checkpoint/rollback system |
| v0.7.0 — Apr 3, 2026 | Profiles (isolated configs), credential pooling, Home Assistant integration |
| v0.8.0 — Apr 8, 2026 Intelligence |
Background process auto-notifications, live model switching (/model), Google AI Studio (Gemini) native provider, MCP OAuth 2.1, approval buttons on Slack and Telegram, Fast Mode for OpenAI/Anthropic |
| v0.9.0 — Apr 13, 2026 Everywhere |
Local web dashboard, iMessage via BlueBubbles, WeChat/WeCom support, Termux/Android support, 16 messaging platforms, Fast Mode expansion, pluggable context engine, comprehensive security hardening, hermes backup and import |
| v0.10.0 — Apr 16, 2026 Tool Gateway |
Nous Tool Gateway — bundled web search, image generation, TTS, and browser automation for Portal subscribers with zero extra API keys |
| v0.11.0 — Apr 23, 2026 | Kanban multi-agent work queues, extended security pass, performance optimizations |
| v0.12.0 — Apr 30, 2026 | Automation blueprints, expanded MCP catalog installs, improved compression |
| v0.13.0 — May 7, 2026 Tenacity |
Reliability overhaul, skill curator background maintenance, multi-provider failover improvements |
| v0.14.0 — May 16, 2026 | Voice mode improvements, additional TTS/STT providers, desktop app groundwork |
| v0.15.0 — May 28, 2026 Velocity |
Major performance release — faster startup, leaner tool schemas, optimized context pipeline |
| v0.15.1–0.15.2 — May 29, 2026 Patch |
Bug fixes and stability improvements following the Velocity release |
| v0.16.0 — Jun 5, 2026 Surface |
Native desktop app (macOS/Linux/Windows) with one-click install, in-app self-update, drag-and-drop, model picker; web dashboard admin panel; Nous Portal quick setup; fuzzy model search; /undo for multi-turn rollback |
| v0.17.0 — Jun 19, 2026 Reach |
iMessage via Photon (no Mac relay), Raft agent network, background/async subagents, image-to-image editing, Automation Blueprints, Cursor Composer model via xAI Grok, full dashboard profile builder, Skills Hub rehaul, atomic batch memory operations, WhatsApp Business Cloud API, Telegram rich text (Bot API 10.1) |
| v0.18.0 — Jul 1, 2026 Judgment |
Mixture-of-Agents as first-class model (named ensembles), live MoA reasoning streaming, self-verifying agent (/goal completion contracts), /learn command for skill creation, /journey learning timeline, background fan-out delegation, desktop coding projects with git worktree management, gateway scale-to-zero and drain coordination, Google Vertex AI provider, P0/P1 clean sweep (700+ issues resolved) |
What People Are Doing With It
Hermes Agent has found traction across a wide spectrum of use cases, from daily software development to home automation. Here are some real-world examples from the community and ecosystem:
Software Development and Code Migration
Teams are using Hermes for autonomous code migration, code review, and multi-step development workflows. The computer-use-linux community project provides Linux desktop control via AT-SPI accessibility trees, letting Hermes drive IDEs and GUI applications directly. The Kanban system enables multi-agent coordination where separate Hermes profiles work on different parts of a codebase simultaneously.
Messaging-Based AI Assistant
The most common use case: running Hermes as a personal AI assistant accessible from Telegram, Discord, or Slack. The gateway supports 16+ platforms, meaning you get full agent capabilities (terminal access, file operations, web search) from your phone. Voice memo transcription is built in, and cron scheduling lets Hermes proactively send daily briefings or reports. The messaging gateway documentation covers the full setup.
Automated Content Pipelines
Hermes’s cron scheduler and skill system make it well-suited for automated content workflows — daily blog post generation, news aggregation, social media scheduling, and RSS feed monitoring. The Skills Hub hosts community-created skills for WordPress publishing, YouTube transcript processing, and tech news aggregation. The Kanban multi-agent board enables chained pipelines where one agent collects data and another processes it.
Research and Data Analysis
The delegate_task system allows fanning out parallel research subagents — “research these five competitors simultaneously” is a single command. Background subagents run without blocking your conversation, and results consolidate into a clean summary. Combined with web search, browser automation, and Python sandbox execution, Hermes functions as a research workstation. The built-in session search (FTS5-backed) means you can find insights from past conversations months later.
Smart Home Automation
First-class Home Assistant integration lets Hermes control lights, switches, sensors, and automations. Combined with the OpenHue skill for Philips Hue control, users have built voice-controlled and schedule-driven home automation systems accessible from any messaging platform.
ML/AI Engineering Workflows
The Hermes Agent skills catalog includes deep expertise for ML engineering: LLM fine-tuning with Axolotl and TRL, GGUF quantization for local inference, RAG pipeline construction with DSPy, and GPU deployment on Modal serverless infrastructure. The GitHub repository includes skills for stable diffusion image generation, Whisper speech recognition, and CLIP multimodal workflows.
WeChat and Chinese Ecosystem
Community-built bridges extend Hermes into the Chinese messaging ecosystem. The HermesClaw project enables running Hermes Agent on WeChat accounts, while native WeCom (enterprise WeChat) and DingTalk adapters ship with the framework. Feishu/Lark document and drive tools are built in.
Getting Started: Installation Guide
1. Install
Linux, macOS, WSL2:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bashWindows (native, PowerShell):
iex (irm https://hermes-agent.nousresearch.com/install.ps1)The installer handles everything: uv, Python 3.11, Node.js, ripgrep, ffmpeg, and a portable Git Bash on Windows. No manual dependency collection needed.
2. Configure Your Model
hermes modelThis launches an interactive picker. Choose from Nous Portal (300+ models with bundled tools), OpenRouter, Anthropic, OpenAI, Google Gemini, DeepSeek, xAI, or any custom endpoint. If you want everything under one subscription, use:
hermes setup --portalThis logs in via OAuth, sets Nous as your provider, and turns on the Tool Gateway (web search, image generation, TTS, browser automation — no extra API keys).
3. Start Chatting
hermesThis opens the interactive CLI with full TUI — multiline editing, slash commands, streaming tool output, and conversation history. For a single one-shot query:
hermes chat -q "What is the capital of France?"4. Connect Messaging Platforms
hermes gateway setup
hermes gateway startNow you can talk to Hermes from your phone. The same agent, same tools, same memory — accessible from anywhere.
5. Explore Further
hermes tools— Enable/disable toolsets (terminal, web, browser, vision, image generation, etc.)hermes skills browse— Browse and install community skills from the Skills Hubhermes cron list— Set up scheduled automationshermes doctor— Diagnose any configuration issues/help— See all slash commands in a session
Community and Resources
- GitHub: NousResearch/hermes-agent — Source code, issues, contributing guide
- Documentation: hermes-agent.nousresearch.com/docs — Complete guides and references
- Discord: discord.gg/NousResearch — Community chat and support
- Skills Hub: agentskills.io — Browse and publish skills
- Nous Portal: portal.nousresearch.com — Managed models and tool gateway
Hermes Agent is actively developed with frequent releases — v0.18.0 (July 2026) closed 100% of P0 and P1 issues in the repo. The project moves fast, with 370+ community contributors and a core team that ships named releases every 2-3 weeks. If you’ve been looking for an AI agent that actually remembers you, works across all your tools, and gets better the more you use it — give Hermes a try.