Description
Hermes Agent is a free, open-source, self-hosted AI agent built by Nous Research, the lab behind the Hermes, Nomos, and Psyche model families. Its defining feature is a closed learning loop: it creates its own skills from experience, improves them the more it’s used, and keeps a persistent, cross-session memory of who you are and what you’re working on.
Key Features
- Self-Improving Skills: Autonomously creates and refines skills from experience, synced with the open agentskills.io standard.
- Persistent Memory: Agent-curated memory with cross-session recall and an evolving model of the user.
- Voice Mode: Full voice interaction across the CLI and messaging platforms, with 10 native text-to-speech providers.
- Sub-Agents: Spawns isolated sub-agents for parallel workstreams, each reporting back on completion.
- Built-In Cron Scheduler: Describe a job in plain language and let Hermes run it unattended, like daily reports or nightly backups.
- MCP Integration: Connects to any MCP server to reach external tools, databases, and internal APIs.
How It Works
A user installs Hermes with a single command and connects it to a terminal backend of their choice — local, Docker, SSH, or a serverless option like Daytona or Modal that hibernates when idle. From there, the agent is reachable from the CLI or from 20+ messaging platforms including Telegram, Discord, Slack, and WhatsApp, all through one gateway. As Hermes handles real tasks, it builds and refines its own skills, stores what it learns in persistent memory, and can hand off long-running jobs to isolated sub-agents or a built-in cron scheduler, so work keeps happening even when the user isn’t actively chatting with it.
Technical Architecture & Overview
- Core Engine: A closed learning loop combining agent-curated memory, autonomous skill creation, and FTS5 cross-session recall with LLM summarization.
- Deployment: Runs on local hardware, Docker, SSH, or serverless backends like Daytona and Modal; supports Linux, macOS, and WSL2.
- API Surface: MCP server support, IDE integration via ACP (VS Code, Zed, JetBrains), and a proxy mode that exposes an existing Claude Pro, ChatGPT Pro, or SuperGrok session as a local endpoint.
- Known Limits: Native Windows support is experimental and requires WSL2; as newer open-source software, some integrations are still actively evolving between releases.
Pros & Cons
| Pros | Cons |
|---|---|
| Free, open source, and self-hosted, with no telemetry or cloud lock-in. | Native Windows support is still experimental and requires installing WSL2. |
| Its self-improving skills and persistent memory genuinely set it apart from stateless chatbots. | Setting it up requires comfort with a terminal, even though day-to-day use happens over chat. |
| Can run on serverless infrastructure that costs nearly nothing when idle, keeping hosting cheap. | Users still pay their own AI provider’s API costs separately from the free software. |
| Built by Nous Research, an established open-source AI lab with real model-training credibility. | As a fast-shipping project (multiple releases in mid-2026 alone), features and defaults change quickly. |
Pricing
| Plan | Price |
|---|---|
| Hermes Agent software | Free (open source, MIT licensed) |
Platform Availability
Linux | macOS | WSL2 (Windows) | Telegram | Discord | Slack | WhatsApp | CLI
Best For
Developers wanting an agent that improves over time | Power users comfortable self-hosting
Similar Tools in This Directory
Also worth comparing: OpenClaw, free self-hosted AI personal assistant | Manus AI, autonomous agent with its own virtual computer | Superhuman, AI-powered email client built for speed
Visit Official Website
Frequently Asked Questions
What is Hermes Agent?
Hermes Agent is a free, open-source, self-hosted AI agent from Nous Research that creates its own skills from experience and remembers across sessions.
Is Hermes Agent free?
Yes, it’s free and open source under the MIT License; users only pay for the API costs of whichever AI model provider they connect.
What makes Hermes Agent self-improving?
It has a built-in learning loop that creates new skills from experience, improves them during use, and persists what it learns across sessions.
Can I switch from OpenClaw to Hermes Agent?
Yes, the setup wizard can automatically detect an existing OpenClaw installation and offers to migrate settings, memories, skills, and API keys.
Which messaging apps does Hermes Agent support?
It supports 20+ platforms including Telegram, Discord, Slack, WhatsApp, Signal, Matrix, Microsoft Teams, and Google Chat from one gateway.
