DEV.co
AI Frameworks · NousResearch

hermes-agent

Hermes Agent is a Python-based AI agent framework from Nous Research that runs locally or in the cloud with built-in memory, skill learning, and multi-platform chat integrations (Telegram, Discord, Slack, etc.). It supports any LLM provider and includes scheduling, tool delegation, and session persistence features.

Source: GitHub — github.com/NousResearch/hermes-agent
211k
GitHub stars
38.7k
Forks
Python
Primary language
MIT
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
RepositoryNousResearch/hermes-agent
OwnerNousResearch
Primary languagePython
LicenseMIT — OSI-approved
Stars211k
Forks38.7k
Open issues26.6k
Latest releasev2026.7.7 (2026-07-08)
Last updated2026-07-08
Sourcehttps://github.com/NousResearch/hermes-agent

What hermes-agent is

Self-hosted AI agent written in Python with autonomous skill creation, FTS5-backed session search, LLM-summarized cross-session recall, cron scheduling, RPC-based tool delegation, and support for six terminal backends (local, Docker, SSH, Singularity, Modal, Daytona). Provider-agnostic LLM routing via Nous Portal, OpenRouter, OpenAI, or custom endpoints.

Quickstart

Get the hermes-agent source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/NousResearch/hermes-agent.gitcd hermes-agent# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

Always-on automation with cloud persistence

Deploy on $5 VPS or serverless (Modal/Daytona) for scheduled tasks, nightly backups, and daily reports running unattended without laptop dependency.

Multi-channel conversational AI

Unified agent accessible via CLI, Telegram, Discord, Slack, WhatsApp, Signal with conversation continuity and voice transcription across platforms.

Agentic research and model training

Batch trajectory generation and compression for training tool-calling models; compatible with agentskills.io open standard for reproducible agent workflows.

Implementation considerations

  • Requires manual setup of LLM provider credentials (Nous Portal, OpenAI, OpenRouter, or custom endpoint); no bundled defaults simplify deployment.
  • Multi-platform gateway (Telegram, Discord, etc.) needs per-platform OAuth/bot tokens and separate API key management—Nous Portal Tool Gateway consolidates but adds another auth layer.
  • Agent memory and skill persistence use local SQLite/file storage by default; cloud deployments (Modal, Daytona) handle persistence but require provider account setup.
  • Six terminal backends (Docker, SSH, Singularity, Modal, Daytona) support different infra but each has distinct dependency and auth requirements.
  • Streaming tool output and interrupt-and-redirect rely on TUI—headless/API-only use cases need custom gateway integration.

When to avoid it — and what to weigh

  • Strict air-gapped deployments — Multi-platform integrations (Telegram, Discord, etc.) and cloud provider support assume outbound network access; requires careful network policy review.
  • Minimal operational overhead required — Configuration spans model selection, tool enablement, provider auth, and platform gateways—setup wizard helps but adds management surface vs. simple API wrappers.
  • Binary portability without dependency management — Requires Python 3.11, Node.js, ripgrep, ffmpeg, and Git; installer bundles dependencies but native Windows installs require separate antivirus whitelisting due to uv.exe false positives.
  • Proprietary or tightly controlled model access — Provider-agnostic design means org policies restricting model choice, API routing, or data residency will require custom configuration or fork.

License & commercial use

MIT License—permissive, allows commercial use, modification, and redistribution with attribution. No GPL copyleft restrictions; can be used in proprietary products.

MIT is a standard permissive OSI license explicitly permitting commercial use. Hermes Agent itself is free to deploy and modify. Commercial use of third-party integrations (Nous Portal, OpenAI models, Firecrawl, etc.) depends on those providers' terms—review each service agreement separately. No guarantee Nous Research provides commercial support or SLAs without explicit contract.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityClear
Deployment complexityHigh
DEV.co fitGood
Assessment confidenceHigh
Security considerations

Self-hosted or cloud-deployed; no security audit or formal hardening guidance documented. Considerations: (1) Agent memory stored locally (SQLite/files) or on third-party cloud providers (Modal, Daytona)—no encryption-at-rest details given. (2) Multi-platform gateway (Telegram, Discord, etc.) exposes agent API to external services—authentication and rate-limiting implementation Unknown. (3) Tool RPC and subagent delegation span Python scripts and cloud terminals (SSH, Singularity, Docker, Modal)—process isolation and privilege escalation risks not documented. (4) LLM provider keys, OAuth tokens, and tool gateway credentials stored in config—no guidance on secrets management or rotation. (5) Session search and cross-session recall via LLM summarization may leak sensitive data if model is accessed by untrusted party. Recommend: security review before handling sensitive workloads, isolation testing for subagents, and secrets audit.

Alternatives to consider

OpenAI Assistants API / GPTs

Hosted, managed agent framework with no self-hosting overhead. Trade-off: vendor lock-in, limited multi-platform integration, no offline operation, and no skill/memory persistence across conversations.

Anthropic Claude Agents (with tool use)

Native Claude with function calling and extended thinking. Trade-off: no persistent memory/skills, no scheduled automation, and limited platform integrations without custom wrapper.

LangChain / LlamaIndex + FastAPI

Build custom agent from components. Trade-off: higher engineering lift, no built-in TUI/multi-platform gateway, requires separate deployment and memory management.

Software development agency

Build on hermes-agent with DEV.co software developers

Evaluate Hermes against your specific security, operational, and integration requirements. Contact Nous Research for commercial support and SLA discussions.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

hermes-agent FAQ

Can I run Hermes offline?
Terminal/TUI works offline, but LLM calls require network. Most platforms (Telegram, Discord) inherently require internet. Local LLM support not explicitly documented but provider-agnostic design suggests custom endpoints possible.
What happens to my conversation history and skills?
Local deployments store memory/skills in SQLite/files under `~/.hermes`. Cloud deployments (Modal, Daytona) persist via provider storage. Migration between backends or provider changes Unknown—review config docs.
Do I need to manage multiple API keys?
Yes, unless using Nous Portal. Direct integrations require: LLM provider key, web search key, image gen key, TTS key, etc. Nous Portal consolidates these under one subscription.
Is Hermes suitable for production / business-critical tasks?
Unknown—no SLA, uptime guarantee, or security audit publicly documented. MIT license includes no liability clause. Recommend testing on non-critical workloads first, and contact Nous Research for commercial support terms if required.

Work with a software development agency

Need help beyond evaluating hermes-agent? DEV.co is a software development agency offering software development services and web development for teams of every size. Our software developers and web developers build custom software, web applications, APIs, and ai frameworks integrations — and maintain them long-term.

Ready to deploy a self-hosted AI agent?

Evaluate Hermes against your specific security, operational, and integration requirements. Contact Nous Research for commercial support and SLA discussions.