Hermes Agent
The self-improving AI agent by Nous Research. The only agent with a built-in learning loop — it creates skills from experience, persists memory across sessions, and deepens its understanding of you over time.
What is Hermes Agent?
Hermes Agent is an open-source AI agent from Nous Research, distinguished by its unique built-in learning loop. Rather than starting fresh every session, Hermes autonomously creates reusable skills from the tasks you give it, refines those skills based on outcomes, and consolidates insights into long-term memory — producing an agent that genuinely improves at your specific workflows over time.
It pairs a full-featured terminal UI with 40+ built-in tools, a cron scheduler, parallel subagent spawning, and broad messaging-channel support. Hermes runs on a $5 VPS, Docker, serverless infrastructure (Modal, Daytona), or a GPU cluster — and supports bring-your-own-key (BYOK) across nine providers including OpenRouter (200+ models), OpenAI, and NVIDIA NIM.
What sets Hermes apart
Built-in learning loop
Autonomously creates reusable skills from your tasks and refines them with each use — no manual prompting required.
Persistent cross-session memory
Consolidates knowledge into long-term memory and builds a deepening user model that compounds across every conversation.
Full terminal UI
Multiline input, slash-command autocomplete, and a rich TUI — comfortable for power users who live in the terminal.
Built-in cron scheduler
Schedule recurring tasks directly inside the agent — no external cron manager or orchestration layer needed.
Parallel subagent spawning
Spawn multiple subagents in parallel to tackle independent parts of a complex task simultaneously.
40+ tools + MCP support
Web search, file operations, shell execution, browser control, and full MCP server support for extending capabilities.
RPC Python scripting
Expose custom Python functions to the agent via RPC so you can wire in any business logic or data source.
Command approval & security controls
Configurable approval gates let you review and confirm sensitive operations before Hermes executes them.
Multi-platform messaging
Receive and respond across Telegram, Discord, Slack, WhatsApp, Signal, and Email alongside the built-in CLI.
Channels & providers
Messaging channels
LLM providers (BYOK)
Bring your own API key — Hermes supports custom endpoints too.
Run it your way
Self-host
Install Hermes on any server, Docker container, or serverless platform. You manage updates, backups, and monitoring.
Requirements
- Python 3.11+
- Node.js
- ripgrep, ffmpeg, git
- Your LLM provider API key (BYOK)
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bashWant production setup with HTTPS and process management? Read the full self-host guide →
Managed by AgentOcean
Launch managed Hermes on a dedicated VPS in ~30 seconds — updates, backups, HTTPS and monitoring handled for you.
- Live in ~30 seconds
- Dedicated VPS — isolated from other users
- Automatic updates & daily backups
- HTTPS + custom domain support
- 24/7 uptime monitoring
- Your API key never touches our servers
Also available: a cheaper self-managed optimized-VPS option — same pre-tuned server, you handle the agent.
View managed Hermes hostingFrequently asked questions
What is Hermes Agent?
Hermes Agent is an open-source AI agent by Nous Research — the only agent with a built-in learning loop. It creates new skills from experience, refines them during use, consolidates knowledge into long-term memory, searches its own past conversations, and builds a deepening model of you across sessions.
What is Hermes Agent's learning loop and self-improvement?
Hermes autonomously creates reusable skills from the tasks you give it, refines those skills based on outcomes, and consolidates insights into long-term memory. It also searches past conversations (via LLM-generated summaries) so context compounds over time. The result is an agent that genuinely improves at your specific workflows the more you use it.
Who makes Hermes Agent?
Hermes Agent is made by Nous Research and is released as open source under the MIT license. The repository is at github.com/NousResearch/hermes-agent.
Where can Hermes Agent run?
Hermes runs on a $5 VPS, via Docker, on any SSH server, serverless (Modal or Daytona with near-zero idle cost), or a GPU cluster. Requirements: Python 3.11+, Node.js, ripgrep, ffmpeg, and git.
What does it cost to run Hermes Agent?
Hermes Agent is free and open source (MIT). You bring your own API key (BYOK) from Nous Portal, OpenRouter, OpenAI, NVIDIA NIM, Hugging Face, or other supported providers. Infrastructure costs start at around $5/mo for a small VPS, or self-managed hosting from $9.99/mo and fully managed from $13.99/mo via AgentOcean.
Can AgentOcean host Hermes Agent for me?
Yes. AgentOcean offers fully managed Hermes Agent hosting on a dedicated VPS. We handle provisioning, HTTPS, daily backups, updates, and 24/7 monitoring — your agent is live in about 30 seconds. A cheaper self-managed optimized-VPS option is also available.
Ready to launch your Hermes agent?
AgentOcean handles the VPS, HTTPS, daily backups, and automatic updates. Your Hermes agent is live in about 30 seconds — bring your own API key and start building.