Open-source, self-hosted AI agent that gets smarter the longer you use it.
Runs on your laptop, VPS or Docker host,
remembers Memorylessons, distills Skillsknow-how,
and stays in touch via Telegram, Discord, Slack, WhatsApp and more.
curl -fsSL https://res1.hermesagent.org.cn/install.sh | bashirm https://res1.hermesagent.org.cn/install.ps1 | iexHermes Agent runs on your laptop, VPS, Docker, SSH or cloud dev environment. It is not tied to a single IDE and never locks your workflow inside a web chat box.
See install optionsOverviewIt remembers your projects, preferences and work patterns across sessions, and distills solved problems into reusable Skills you can compose later.
See the memory systemOverviewSupports MCP, terminal, files, browser, image, TTS and more, with cron scheduling for briefings, backups, sweeps and reminders.
See automationOverviewStay in touch via Telegram, Discord, Slack, WhatsApp, Signal, Email and many more — your agent is always reachable.
See integrationsOverviewDesigned for jobs that span days, weeks and months — not just one chat. Sessions, profiles, checkpoints and rollbacks are first-class concepts.
Get startedOverviewWorks with Claude, GPT, Gemini, Qwen, GLM, Kimi, MiniMax, OpenAI-compatible endpoints and local models. Pick whatever fits your network and workflow.
See model configurationcurl -fsSL https://hermes-agent.nousresearch.com/install.sh | bashirm https://hermes-agent.nousresearch.com/install.ps1 | iexFull troubleshooting guide: see the install docs
hermes setup
hermes modelhermeshermes gateway setup
hermes gatewayHermes Agent is an open-source, self-hosted AI agent with long-term memory and Skills. It runs in your terminal for complex tasks, and stays online through messaging gateways like Telegram, Discord and Slack.
Quickstart ↗IDE assistants are great at instant collaboration inside the current window. Hermes Agent emphasises long-term context, cross-session memory, reusable Skills, MCP integrations and multi-platform automation — built for jobs that keep running.
See long-term memory ↗Start with the install guide and the Quickstart. From there, configure a model, then add tools and a messaging gateway as your workflow grows.
See install guide ↗Through long-term memory, session search, skill distillation and user profiles, Hermes turns each task into reusable expertise.
See the memory systemHermesRun it on a VPS, Docker, SSH, Modal, Daytona or local box — and talk to it through Telegram, Discord, Slack, WhatsApp, Signal, Email and CLI.
See integrationsHermesBuilt-in cron scheduling: briefings, backups, sweeps, scrapes and reminders — all configurable in plain English and delivered to the platforms you live in.
See the automation guideHermesDistill complex tasks into reusable skills, compatible with the agentskills.io open format. Great for teams and community contribution.
See skillsHermes40+ built-in tools, toolset profiles, command approval, container isolation, browser control, vision, image generation, TTS and multi-model inference.
See the tool referenceHermesBulk trajectory generation, compression, Atropos RL environments and data workflows for next-generation tool-calling models.
See RL training