Hermes Agent Documentation

Upstream Hermes Docs

Hermes Agent is developed by NousResearch. The canonical documentation is hosted at hermes-agent.nousresearch.com/docs. Key sections:

• Installation — Getting started on all platforms
• Docker Guide — Running Hermes in a container
• API Server — OpenAI-compatible API on port 8642
• Web Dashboard — Dashboard UI on port 9119
• Skills — Built-in, optional, and community skills
• Skills Hub — 652 skills (72 built-in, 59 optional, 521 community)
• CLI Reference — All hermes subcommands
• MCP with Hermes — Configuring MCP servers in config.yaml
• NousResearch/hermes-agent on GitHub — Source code, issues, and release tags

Configuration is driven by a config.yaml file and environment variables in ~/.hermes/.env.

Quick-Reference: Confirmed Capabilities

The following list is drawn entirely from upstream source inspection of Hermes Agent v2026.4.16. Do not rely on it for later versions — check the repository for changes.

• OpenAI-compatible API server — default port 8642. Endpoints include /v1/chat/completions, /v1/models,/v1/runs, and /health. Enable by setting API_SERVER_ENABLED=true in ~/.hermes/.env. Session continuity via previous_response_id, the conversation parameter, and the Runs API.
• Dashboard — separate command: hermes dashboard, port 9119, localhost only, no built-in auth. Optional install: pip install hermes-agent[web]. Covers Analytics, Config, Cron, Env, Logs, Sessions, Skills, and Status.
• Image generation — FAL.ai (env: FAL_KEY,FAL_IMAGE_MODEL) plus image skills in the Skills Hub (pixel-art, p5js, and others).
• Web search — four first-class providers: Tavily, Exa, Parallel, and Firecrawl. Auto-selected by which API key is set. DuckDuckGo available as an optional skill (no key required) and activates automatically when no paid search key is configured.
• Vision inputs — centralized aux router supporting OpenRouter, Nous, Codex, native Anthropic, and custom OpenAI-compatible endpoints.
• Audio transcription — via tools/transcription_tools.py.
• Shell execution — local, Docker, Modal, SSH, Singularity, and Daytona environments. Foreground and background modes.
• Approval policy — approvals.mode: manual, smart, or off. Command allowlist with once / session / always scope. Configured at root config level (not under tools).
• Agent personalities — 14 presets (helpful, concise, technical, creative, teacher, kawaii, pirate, and more). CLI-level config.
• Pairing — hermes pairing approve <platform> <code> CLI. Credentials stored at /opt/data/platforms/pairing/.
• Skills Hub — 652 skills (72 built-in, 59 optional, 521 community). Includes media generation via skills: video (ascii-video, manim-video), music (heartmula, audiocraft-audio-generation), and image (pixel-art, p5js). See the Skills Hub for the full catalogue.
• MCP — configure via mcp_servers: in config.yaml; CLI hermes mcp available.

Install and Setup

New to Hermes? These guides cover getting it running from scratch.

• How to Install Hermes Agent — Step-by-step installation for all platforms
• Hermes Agent with Docker Compose — Production-ready Compose setup with volumes and env file
• Hermes Agent on Windows — Windows setup via WSL2 or Docker Desktop

Connecting Channels

Hermes supports the same platform connection flow as OpenClaw. Pairing, connect, and disconnect steps are consistent across all channels.

• Deploy Hermes Agent — Full deployment walkthrough including first channel connection
• Hermes Agent Web UI Guide — Using the dashboard on port 9119 to configure and monitor your agent

Per-channel setup (Telegram, Discord, Slack, and others) mirrors the OpenClaw connect flow. Each platform has its own config fields; see the platform-specific pages in the upstream docs.

Models and Providers

Hermes supports multiple model providers configured via environment variables and config.yaml. Provider selection is determined at startup; switching providers requires a restart. See the upstream README for the full list of supported providers and the required environment variable for each.

• Hermes Agent + Ollama — Run Hermes with local models via Ollama

Skills, Tools, and Memory

Hermes skills extend agent capabilities. Optional skills can be enabled per instance. Memory persists across sessions when configured.

• Hermes Agent Skills Guide — Installing, enabling, and writing custom skills
• Hermes Agent Memory — How session memory and context persistence works

Platform Comparison and Hosting

Evaluating Hermes vs. OpenClaw, or deciding where to host? These pages cover both.

• Hermes Agent vs. OpenClaw — Side-by-side feature and capability comparison
• Self-Deploy vs. OpenClaw Launch — What you get from managed hosting vs. running it yourself
• Hermes Agent Hosting Options — VPS, Docker, cloud, and managed hosting compared
• Hermes Agent on Hostinger — Setup guide for Hostinger VPS deployments

Open Source and Community

• Hermes Agent GitHub Guide — How to browse the source, open issues, and contribute
• What is Hermes Agent? — Plain-language overview of what Hermes is and who it is for

Run Hermes Without Reading Any More Docs

If you would rather skip the setup and get straight to using Hermes, OpenClaw Launch offers one-click deployment of a fully managed Hermes instance. TLS, container management, updates, and channel pairing are all handled for you. Your instance is ready in seconds — no config.yaml editing, no port forwarding, no Docker commands.

Frequently Asked Questions

Is there official Hermes Agent documentation?

Yes. The canonical docs are at hermes-agent.nousresearch.com/docs, covering installation, Docker setup, the API server, the web dashboard, skills, the CLI, and MCP integration. The NousResearch/hermes-agent repository on GitHub is also a reference for source-level details. OpenClaw Launch maintains this guide hub as a supplement with hosted deployment walkthroughs and hosting context.

Where do I report a Hermes docs bug or missing topic?

Open an issue on the GitHub Issues tracker. For issues specific to this guide hub, email [email protected].

Does OpenClaw Launch maintain its own Hermes docs?

Yes. The guides linked on this page are maintained by OpenClaw Launch and tested against current Hermes versions. They cover topics that the upstream README does not address in depth — hosted deployments, channel pairing walkthroughs, and comparison with other agent frameworks. For configuration field definitions and API specs, always defer to the upstream repository.

Is Hermes Agent documented in other languages?

The upstream repository documentation is in English only. The guides on this page are also English only. If you need help in another language, the GitHub Issues tracker and community channels are the best places to ask.

Does Hermes support video or music generation?

Yes, via the Skills Hub. Media generation skills include video (ascii-video, manim-video), music (heartmula, audiocraft-audio-generation), and image (pixel-art, p5js), among 652 total skills. See the Skills Hub for the full catalogue.