SkillFoxx

OpenHermit CLI

OpenHermit runs a fleet of AI agents as production services: Postgres state, sandboxes, Telegram, Discord and Slack channels

  • 90 ★ on the repository
  • Last commit
  • MIT
CLI

Broad access

Why this level

  • The platform runs agents that execute code in sandboxes and stores secrets, memory and user sessions in PostgreSQL
  • Agents answer in Telegram, Discord and Slack and run on a schedule unattended, and an `--all` command changes the behavior of the whole fleet
All reasons and checks

hcf-studios/openhermitAuthor:HCF-STUDIOS

Install

Works in

In your terminal, with SkillFoxx CLI

npx skillfoxx add cli/openhermit

Detects the agents on your machine, shows the access level and pins the version.

Other ways to install

Access and checks

How we review

Broad accessThe tool can change data, run commands or handle secrets.

  • The platform runs agents that execute code in sandboxes and stores secrets, memory and user sessions in PostgreSQL
  • Agents answer in Telegram, Discord and Slack and run on a schedule unattended, and an `--all` command changes the behavior of the whole fleet

This is third-party code. Review the repository files before installing.

What it does

An open-source platform for running AI agents as persistent services. The core idea: internal state is separated from external state. Sessions, memories, instructions, skills, MCP servers, schedules, secrets and users live in shared PostgreSQL scoped by `agent_id`, while an agent's working files sit in its sandbox: its own Docker container or cloud E2B, Daytona or Tenki. One Hono gateway manages all agents and serves an admin UI at `/admin/`. Fleet operations use `hermit` commands: `hermit skills enable ... --all` turns a skill on for every agent, `hermit mcp enable ... --all` attaches an MCP server, `hermit instructions append` adds a rule and `hermit config secrets set` rotates a secret. Channels are Telegram, Discord and Slack, plus packages for Signal, WeChat and WhatsApp. It has schedules, owner, user and guest roles and deep research with a report where every finding cites a source. Install with `npm install -g openhermit`, then `hermit setup` and `hermit gateway start`.

Who it is for. Teams that run an internal agent platform, a SaaS with an agent per customer or a swarm of roles and do not want to manage files on every machine.

Good fit when

  • A rule must reach every agent with `hermit instructions append rules "..." --all`
  • A skill or MCP server must be turned on across the whole fleet and checked in one place
  • Agents need E2B or Daytona sandboxes and shared state in PostgreSQL
  • A bot must answer in Telegram, Discord and Slack with settings changed without a restart

Not a fit when

  • You are the only user on one machine: `CLAUDE.md` files and a local CLI agent are enough
  • You cannot run PostgreSQL and sandbox containers
  • You only need a terminal coding agent with no channels or fleet

Example request

Start OpenHermit, add an agent with a Telegram channel and enable the standup-digest skill across the team

Limitations

  • Node.js, PostgreSQL and a chosen sandbox backend are required: Docker, E2B, Daytona or Tenki.
  • Models need provider keys, for example OpenAI or Anthropic, and secrets are set with `hermit config secrets set`.
  • Deep research spends many tokens and hits quotas on free model tiers, and the documentation gives worst-case price estimates by depth.
  • Local models are workable from 27 billion parameters, while 7 to 8 billion models often fail answer-format validation.

How to disable. Stop the gateway with `hermit gateway stop`, uninstall the package with `npm uninstall -g openhermit` and, if needed, remove the PostgreSQL database and agent sandboxes.

FAQ

How do I install OpenHermit?

Run `npm install -g openhermit`, then `hermit setup` for first-time setup and `hermit gateway start` to start the gateway. The admin UI opens at `/admin/` on the gateway.

Where is an agent's state stored?

Internal state lives in shared PostgreSQL by `agent_id`, and working files live in the agent sandbox: Docker, E2B, Daytona or Tenki.

Which channels are supported?

Telegram, Discord and Slack are built in, and Signal, WeChat and WhatsApp install as packages. A CLI and a web UI also exist, and channels are enabled and reconfigured at runtime.

How do I roll a skill out to all agents?

Run `hermit skills enable <skill> --all`. For an MCP server use `hermit mcp enable <server> --all`.

Can it run on local models?

Yes, any OpenAI-compatible server works: Ollama, LM Studio, llama.cpp or vLLM. Research needs a model from 27 billion parameters and a context of 64 thousand tokens or more.

Discussion

No questions or reviews yet.

CLI tools

hermitd runs an always-on Claude Code agent on your machine: tasks from chat, scheduled checks and saved state after a restart.

PluginBroad access84

A self-hosted AgentOS for one workspace: plan-mode chat, private knowledge base, MCP tools, skills, memory and a sandbox

CLIBroad access631

oah links 80-plus skills, five roles and /commit, /review, /test-tdd commands into Claude Code, Cursor, Codex and other agent folders

CLIExtended access997
Editors’ pick

A self-improving agent from Nous Research with a TUI, messaging gateway, cron jobs and skills it writes itself

CLIBroad access262.9K

A self-hosted control center for coding agents: OpenHands, Claude Code, Codex and Gemini on local and remote backends

CLIBroad access93.3K

A desktop app that runs Claude Code, Codex, Cursor and OpenCode in parallel, each in its own git worktree, with a CLI and MCP server for control

CLIExtended access1.3K

README in short

The README explains why file-based CLI agent state does not suit teams and describes the idea of splitting internal and external state. Features, installation, a detailed section on cost and model choice for deep research, local models and a command reference follow. Source run commands are given for development.

README badge

Report a problem