←Back to Skill Directory

Agent Framework

30K+ StarsSelf-EvolvingMIT License

Hermes Agent

by Nous Research · hermes-agent.nousresearch.com

Hermes Agent is an open-source AI agent framework built by Nous Research with a closed learning loop at its core. After completing complex tasks, it auto-generates skill documents capturing the strategies used — and self-iterates those skills when it discovers better approaches. The result is an agent that gets measurably better at recurring tasks without manual tuning.

Launched in late February 2026, Hermes Agent reached 30,000+ GitHub stars faster than any other agent framework, driven by its unique combination of autonomous skill evolution, two-layer memory, 5-layer security, and the ability to run the entire stack on a $5/month VPS. It is the first real competitor to OpenClaw in the open-source agent ecosystem — though the two tools solve fundamentally different problems.

30K+
GitHub Stars
Feb 2026 launch
5
Security Layers
auth to injection
6+
Platforms
Telegram, Discord...
$5/mo
Min Hosting
VPS or serverless

Quick Install

curl -sSL https://hermes-agent.nousresearch.com/install | bash

Key Features

Closed Learning Loop

Auto-generates structured skill documents after complex tasks. No manual curation — the agent captures what worked and stores it for reuse.

Self-Iterating Skills

When the agent finds a better method for a known task, it overwrites the existing skill document. Skills evolve through execution, not configuration.

Two-Layer Memory

MEMORY.md for persistent long-term context plus SQLite full-text search over complete conversation and task history for fast retrieval.

5-Layer Security

User auth, command approval gates, container isolation, credential filtering, and injection scanning — all enabled by default.

OpenAI API-Compatible Server

Exposes a standard chat completions endpoint. Connect Open WebUI, custom frontends, or any OpenAI-compatible client directly.

Multi-Model Support

Works with Claude, GPT, Llama, Mistral, and any model served via OpenAI-compatible API. Switch models without changing agent code.

Cheap Self-Hosting

Runs on a $5/month VPS with Docker, or deploy serverlessly on Modal. No GPU required for the agent runtime itself.

Execution Brief

Use this page as a rollout checklist, not just reference text.

Suggest update

Tool Mapping Lens

Organize Tools by Workflow Phase

Catalog-oriented pages work best when users can map discovery, evaluation, and rollout in a clear path instead of reading an undifferentiated list.

  • Define the job-to-be-done first
  • Group tools by stage
  • Prioritize by adoption friction

Actionable Utility Module

Skill Implementation Board

Use this board for Hermes Agent before rollout. Capture inputs, apply one decision rule, execute the checklist, and log outcome.

Input: Objective

Deliver one measurable improvement with hermes agent nous research ai agent framework review

Input: Baseline Window

20-30 minutes

Input: Fallback Window

8-12 minutes

Decision TriggerActionExpected Output
Input: one workflow objective and release owner are definedRun preview execution with fixed acceptance criteria.Go or hold decision backed by repeatable evidence.
Input: output quality below baseline or retries increaseLimit scope, isolate root issue, and rerun controlled test.One confirmed correction path before wider rollout.
Input: checks pass for two consecutive replay windowsPromote to broader traffic with fallback path active.Stable rollout with low operational surprise.

Execution Steps

  1. Record objective, owner, and stop condition.
  2. Execute one controlled preview run.
  3. Measure quality, latency, and correction burden.
  4. Promote only when pass criteria are stable.

Output Template

tool=hermes agent nous research ai agent framework review
objective=
preview_result=pass|fail
primary_metric=
next_step=rollout|patch|hold
Share execution feedback

What Is Hermes Agent?

Hermes Agent is an open-source AI agent framework by Nous Research that introduces a closed learning loop to autonomous task execution. Unlike traditional agent frameworks where capabilities are static unless manually updated, Hermes Agent generates skill documents from its own successful task completions and refines them over time.

The framework solves a fundamental problem in agent systems: knowledge retention. Most agents start from zero context on every session. Hermes Agent maintains a two-layer memory system — MEMORY.md for persistent state and a SQLite database with full-text search for complete conversation history. When a new task arrives, the agent searches its history for relevant skills before starting, dramatically reducing redundant reasoning.

Hermes Agent launched in late February 2026 and accumulated 30,000+ GitHub stars within weeks, making it the fastest-growing agent framework in open-source history. Its appeal comes from combining autonomous skill evolution with practical deployment characteristics: MIT license, $5/month hosting, multi-model support, and production-grade security defaults.

The competitive landscape positions Hermes Agent differently from OpenClaw. OpenClaw is a gateway and routing layer — it manages API keys, rate limits, and model selection across providers. Hermes Agent is an agent runtime — it executes tasks autonomously, learns from execution, and builds a skill library. They are complementary: Hermes can route model calls through OpenClaw for centralized cost and key management.

How to Calculate Better Results with hermes agent nous research ai agent framework review

Install Hermes Agent with: curl -sSL https://hermes-agent.nousresearch.com/install | bash. This sets up the CLI, default configuration, and Docker runtime.

Configure your model provider by setting API keys in the Hermes config file. Hermes supports Claude, GPT, Llama, Mistral, and any OpenAI-compatible endpoint. For self-hosted models, point to your vLLM, Ollama, or LiteLLM server.

Connect a frontend: Hermes Agent exposes an OpenAI API-compatible server. Connect Telegram, Discord, Slack, WhatsApp, CLI, or Open WebUI. For custom integrations, use the standard chat completions endpoint.

Run your first task through the CLI or connected frontend. After completion, check MEMORY.md — you will see the agent has already generated skill documents from the task. On the next similar task, the agent will retrieve and apply those skills automatically.

Treat this page as a decision map. Build a shortlist fast, then run a focused second pass for security, ownership, and operational fit.

When a team keeps one shared selection rubric, tool adoption speeds up because evaluators stop debating criteria every time a new option appears.

Worked Examples

Setting up a research automation workflow

  1. Install Hermes Agent and configure Claude as the model provider
  2. Connect Telegram as the frontend for mobile access
  3. Send first research task: "Analyze the top 10 competitors in the AI code review space"
  4. Agent executes: searches web, reads landing pages, compiles comparison table, writes summary
  5. After completion, agent auto-generates a skill doc: "competitor-analysis-workflow" in MEMORY.md
  6. Next research task uses the learned workflow pattern — 40% fewer model calls, structured output from the start

Outcome: A research agent that learns your preferred analysis format and methodology. Each subsequent research task runs faster and produces more consistent results.

Deploying a personal AI assistant on a $5 VPS

  1. Provision a 1GB VPS ($5/month) and install Docker
  2. Run the Hermes Agent Docker image with Llama 3 via Groq API (free tier)
  3. Connect WhatsApp and Telegram as dual frontends
  4. Configure 5-layer security: set API key, enable command approval for shell access, verify container isolation
  5. Use daily for scheduling, research, and content drafting
  6. After two weeks, the agent has 47 self-generated skill documents covering recurring task patterns

Outcome: A personal AI assistant running 24/7 for $5/month that progressively learns your workflows and preferences without any manual skill configuration.

Frequently Asked Questions

What is Hermes Agent?

Hermes Agent is an open-source AI agent framework created by Nous Research. It features a closed learning loop — after completing complex tasks, it automatically generates skill documents that capture the methods used. On subsequent runs, it retrieves and refines those skills, making the agent progressively more capable over time. It launched in late February 2026 and became the fastest-growing agent framework on GitHub with over 30,000 stars.

How does the closed learning loop work?

When Hermes Agent completes a task that involved multi-step reasoning or tool use, it extracts the successful strategy into a structured skill document stored in MEMORY.md. The next time a similar task appears, it retrieves that skill via SQLite full-text search over its history database. If the agent discovers a better approach, it overwrites the existing skill document — this is the self-iteration mechanism. The result is an agent that learns from its own execution without any manual curation.

How does Hermes Agent compare to OpenClaw?

OpenClaw focuses on gateway routing — it acts as a centralized hub that routes requests to different AI models and manages API keys, rate limits, and cost tracking. Hermes Agent focuses on agent evolution — it runs tasks autonomously, learns from execution, and builds a growing library of self-iterated skills. They solve different problems: OpenClaw is infrastructure for managing multiple AI providers, Hermes is a runtime for building agents that improve themselves. They can be used together — Hermes can route its model calls through an OpenClaw gateway.

What platforms does Hermes Agent support?

Hermes Agent runs as a backend service that exposes an OpenAI API-compatible endpoint. It natively supports Telegram, Discord, Slack, WhatsApp, CLI, and Open WebUI as frontends. Because it implements the OpenAI chat completions API, any client that works with the OpenAI API can connect to Hermes Agent — including custom applications, chatbot frameworks, and developer tools.

How do I install Hermes Agent?

Run: curl -sSL https://hermes-agent.nousresearch.com/install | bash. This installs the CLI and pulls the default configuration. You need an API key for at least one supported model provider (Claude, GPT, Llama, Mistral, etc.). For Docker deployment, use the official image from the repository. The entire stack runs on a $5/month VPS with 1GB RAM, or you can deploy serverlessly on Modal.

What are the security features of Hermes Agent?

Hermes Agent implements five security layers: (1) User authentication with API keys and role-based access, (2) Command approval gates for destructive operations, (3) Container isolation via Docker sandboxing for code execution, (4) Credential filtering that strips secrets from logs and memory, (5) Injection scanning that detects and blocks prompt injection attempts in user inputs and tool outputs. All five layers are enabled by default.

Which AI models work with Hermes Agent?

Hermes Agent supports any model accessible via an OpenAI-compatible API: Claude (via Anthropic API or proxy), GPT-4o and GPT-4.5 (via OpenAI API), Llama 3 and Llama 4 (via Groq, Together, or local inference), Mistral Large (via Mistral API or local), and any model served by vLLM, Ollama, or LiteLLM. You configure the model endpoint and API key in the Hermes config file, and the framework handles the rest.

Missing a better tool match?

Send the exact workflow you are solving and we will prioritize a new comparison or rollout guide.

Submit feedback