Connect Everywhere
Hermes Agent works with your favorite messaging apps, model providers, and tool protocols. Same memory, same skills, reachable from anywhere you already chat.
Messaging Platforms
Hermes can run as a gateway on Telegram, Discord, Slack, WhatsApp, Signal, Email, and the CLI (README). Start a conversation on one platform and pick it up on another — the memory follows you.
Hermes runs a WhatsApp gateway out of the box — the installer fetches Node.js v22 specifically so the WhatsApp bridge can work. Listed in the README alongside Telegram, Discord, and Slack.
settings Run hermes gateway after the install script
Telegram
Listed in the README as a first-class messaging platform. Connect a bot token via the hermes setup or hermes gateway wizard.
settings Paste bot token from @BotFather during hermes setup
Discord
Listed in the README as a first-class messaging platform. Run Hermes as a Discord bot — channels, threads, and DMs all work.
settings Create bot at discord.dev, paste token into hermes setup
Slack
Listed in the README and ships as a dedicated pyproject extra (slack). Install as a Slack workspace app and let your channel talk to Hermes.
settings Create Slack app, install to workspace, add token via hermes setup
Signal
Listed in the README as a first-class messaging platform. Connect via a Signal relay so Hermes can respond to encrypted DMs and groups.
settings Link device during hermes gateway
iMessage
Not officially listed in the Hermes README or docs. The Hermes team positions macOS support as CLI and any connected chat gateway; for iMessage, community members typically bridge via a third-party relay.
settings Community bridge — no official path yet
Microsoft Teams
Not listed in the Hermes README or pyproject extras. No first-party Teams integration is documented upstream.
settings No official integration
Matrix
Ships as a dedicated pyproject extra (matrix). Install with uv pip install -e ".[matrix]" to enable the Matrix gateway.
settings Install the [matrix] extra and configure homeserver
WebChat
Not explicitly documented in the README or install docs. The Hermes CLI is listed as a first-class channel; browser UI is not the primary interface.
settings No dedicated web UI documented
Model Providers
Hermes is model-agnostic. The README lists Nous Portal, OpenRouter, z.ai/GLM, Kimi/Moonshot, MiniMax, OpenAI, and custom endpoints. Anthropic and Mistral ship as first-class dependencies in pyproject.toml.
Tip: Not sure which model to pick? Nous Portal runs Nous Research's own Hermes models and is the house default. OpenRouter routes across 200+ models from one key. Anthropic Claude and OpenAI GPT are shipped as direct dependencies. Run a local model via any OpenAI-compatible endpoint (e.g. Ollama) if you need zero cloud dependencies.
Anthropic
The anthropic Python SDK is a direct dependency in pyproject.toml, so Claude is natively supported as a model provider.
settings Get API key at console.anthropic.com, add via hermes model
OpenAI
The openai Python SDK is a direct dependency in pyproject.toml and is explicitly listed in the README as a supported provider.
settings Get API key at platform.openai.com, add via hermes model
Not named as a first-party provider in the Hermes README or pyproject dependencies. Gemini models can still be reached through an OpenAI-compatible proxy or OpenRouter.
settings Route through OpenRouter or a compatible proxy
xAI
Not named as a first-party provider in the Hermes README or pyproject dependencies. Grok can be reached through OpenRouter or a custom OpenAI-compatible endpoint.
settings Route through OpenRouter or a compatible endpoint
OpenRouter
Explicitly listed in the README as "OpenRouter (200+ models)" — a single API key gives you access to most mainstream model providers.
settings Get API key at openrouter.ai, add via hermes model
Mistral
Ships as a dedicated pyproject extra (mistral). Install with uv pip install -e ".[mistral]" to get the Mistral provider wired up.
settings Install the [mistral] extra, add key via hermes model
DeepSeek
Not named as a first-party provider in the Hermes README or pyproject dependencies. DeepSeek offers an OpenAI-compatible endpoint, so it can be plugged in as a custom endpoint.
settings Configure as custom endpoint in hermes model
Perplexity
Not listed as a first-party provider. Perplexity is a search product rather than a pure model provider; web search in Hermes is handled by the exa-py and firecrawl-py dependencies instead.
settings Use the built-in web search tools
Hugging Face
Not named as a first-party provider. Hugging Face inference endpoints that speak the OpenAI API can be added as custom endpoints.
settings Configure as custom endpoint in hermes model
Local Models
The README says "or your own endpoint" — Ollama exposes an OpenAI-compatible API, so it plugs in cleanly as a local endpoint with zero cloud dependencies.
settings Install Ollama, point hermes model at http://localhost:11434
Sandbox Backends
Hermes supports six terminal sandbox backends according to its README: local, Docker, SSH, Daytona, Singularity, and Modal. Pick one to isolate command execution from your host system.
Cloudflare
Not listed as an official integration. The Hermes team does not ship a Cloudflare Workers deployment path in the upstream repo.
settings No official Cloudflare path
Smart Home
Hermes ships an optional homeassistant extra (pyproject.toml). Control lights, thermostats, locks, scenes, and automations through natural language once you connect a Home Assistant instance.
Home Assistant
Ships as a dedicated pyproject extra (homeassistant). Install with uv pip install -e ".[homeassistant]" to control lights, thermostats, scenes, and automations by chat.
settings Install the [homeassistant] extra, add HA token
Philips Hue
Not a dedicated integration — Philips Hue lights are controlled through the homeassistant extra once your bridge is connected to Home Assistant.
settings Expose Hue through Home Assistant
Developer Tools
Hermes exposes its tools through the open MCP (Model Context Protocol) and ACP (Agent Connect Protocol) standards, so any MCP-compatible editor or tool can talk to it. See docs/user-guide/features/mcp.
GitHub
Not a first-party integration. Connect a community GitHub MCP server through the [mcp] extra — Hermes' docs/user-guide/features/mcp page shows how.
settings Install the [mcp] extra, attach a GitHub MCP server
VS Code
Not a first-party integration. Hermes exposes its tools through MCP, so any editor with MCP client support can drive it. Check the Hermes repo and agentskills.io for community bridges.
settings Use an MCP-compatible VS Code extension
Docker
Docker is one of six terminal sandbox backends listed in the README (local, Docker, SSH, Daytona, Singularity, Modal). Use it to isolate command execution from your host.
settings Set Hermes' sandbox to docker in config
Productivity
Persistent memory (via Honcho), procedural skills from agentskills.io, cron scheduling, and SOUL.md personality config let Hermes remember your projects, preferences, and how you like to work.
Notion
Not a first-party integration. Hermes supports MCP servers — so community Notion MCP servers are the recommended path.
settings Attach a community Notion MCP server
Obsidian
Not a first-party integration. Point Hermes at your vault as plain files and use a community Obsidian MCP server if you need graph-aware access.
settings Attach a community Obsidian MCP server
Trello
Not a first-party integration. No Trello extra or docs page exists upstream.
settings No official integration
Music & Audio
Not officially documented. If you need music control, the community typically wires it up through a custom MCP server or a skill on agentskills.io.
Spotify
Not a first-party integration. No Spotify extra or docs page exists upstream.
settings No official integration
Operating Systems
Hermes runs on Linux, macOS, WSL2 on Windows, and Android via Termux (README). Native Windows is explicitly not supported.
macOS
Installer supports macOS directly — the quick install script detects macOS and runs the standard uv-based flow. The CLI is the primary interface.
settings Run the curl install script
iOS
No native iOS app is documented upstream. Reach Hermes from iOS by connecting any chat gateway (Telegram, Signal, etc.) to an always-on host.
settings Connect Hermes on another host, chat from iOS
Android
The README explicitly supports Android via Termux. The install script auto-detects Termux and uses the tested .[termux] extra instead of .[all].
settings Run the install script inside Termux
Linux
First-class Linux support — the install script is primarily tested on Linux distributions, using uv to manage Python 3.11 and a venv under ~/.hermes/hermes-agent/.
settings Run the curl install script
Need a Custom Integration?
Hermes Agent exposes its tools through MCP and ACP. Build your own integration as a skill or an MCP server — the Nous Research repo is the source of truth.