Index / Hermes Agent · updated Oct 24, 2026
Hermes Agent
NousResearch/hermes-agent · healthy · rank 2 of 50 by stars
A self-improving personal AI agent from Nous Research with a built-in learning loop that creates and refines skills from experience. It runs anywhere from a $5 VPS to serverless, with a polished TUI and multi-platform messaging gateway.
Facts
Repository
Runtime
Posture
Community
Security breakdown
Composite 72 / 100 · how these are scored
higher is safer
higher is safer
higher is safer
higher is safer
higher is riskier
Model access
Read from the repository, not written by a model · 109 files examined
11 providers · runs locally · gateway support
Pinned models
claude-opus-4.6 released 2026-02-04 — 9 months old (from the public model catalogue)
Pin last edited 2 days ago
Evidence
Decision
Why choose Hermes Agent over OpenClaw?
Why choose this
- Built-in closed learning loop with autonomous skill creation and self-improvement
- Native multi-platform messaging gateway (Telegram, Discord, Slack, WhatsApp, Signal)
- Model-agnostic provider switching with no lock-in
Tradeoffs
- Heavier Python runtime footprint than leaner OpenClaw deployments
- Unrestricted shell access raises operational risk without manual Docker isolation
- Rapid issue/PR velocity strains maintainer throughput per ecosystem digests
Best fit
- Users wanting a persistent agent with long-term memory and skill learning
- Self-hosters running agents on cheap VPS or serverless infra
- People who chat via Telegram/Discord/Slack rather than a desktop app
Avoid if
- You need strict sandboxing of agent shell access out of the box
- You want a lightweight, minimal-footprint agent
- You prefer fully offline operation with no external model providers
Strong evidence from an extensive README, very active recent commits, and 19 Reddit threads plus web coverage. Sentiment is mostly positive but mixed on beginner UX and security isolation, so those scores carry moderate uncertainty.
AI layer reviewed Oct 5, 2026 · how this is written
Star activity
256,715 stars today
Overview
Hermes Agent is Nous Research's flagship personal AI agent, positioned as 'the agent that grows with you.' Its defining feature is a closed learning loop: the agent autonomously creates skills after complex tasks, improves those skills during use, nudges itself to persist knowledge, and searches its own past conversations via FTS5 with LLM summarization. It integrates Honcho for dialectic user modeling and is compatible with the agentskills.io open standard, making memory and skill portability first-class concerns.
Architecturally, it is a Python application with a full terminal UI (multiline editing, slash-command autocomplete, streaming tool output) plus a single gateway process that bridges Telegram, Discord, Slack, WhatsApp, and Signal with cross-platform conversation continuity and voice memo transcription. It is model-agnostic — Nous Portal, OpenRouter, OpenAI, or custom endpoints — switchable at runtime via hermes model. Deployment targets range from a $5 VPS to GPU clusters to scale-to-zero serverless, and it ships scheduled automations and a plugin system with a catalog of model-provider and desktop plugins.
Relative to OpenClaw, Hermes competes directly (it even ships hermes claw migrate) and differentiates on its learning loop and messaging-gateway breadth. Recent commits show heavy investment in provider correctness (Anthropic thinking-mode handling), desktop UX polish, and plugin infrastructure. The main caveats are operational: unrestricted shell access means users are advised to run it in Docker, and the project's extremely high inbound issue/PR volume can outpace maintainer throughput.