QM
A team agent you run in your own cloud, reachable through Slack and a web app.
- Stars
- 15.6k
- Forks
- 1.9k
- Licence
- MIT
- Language
- TypeScript
- Last update
- 4 days ago
- Type
- AI model / agent
What it is
QM is a self-hosted agent system meant for a small company or startup, where each employee gets an isolated workspace with its own memory, files, permissions and sandbox. People can talk to the agent privately or together in Slack channels and projects, and also through a web app. You deploy it into your own cloud account and choose which models and harnesses (such as Pi or Claude Code) it uses.
- Search internal notes, email, documents, databases and the web together
- Build internal web apps and publish them to specific people
- Triage an inbox on a schedule, including labels and reply drafts
- Work in an existing repository to run tests, open PRs and check CI
- Track a project in a shared Slack channel and post follow-ups
Great for
- Startups that want a shared AI agent for the whole team in Slack
- Teams that want to keep data and API keys in their own cloud account
- Organizations that want admin control over which models and security settings are allowed
- Developers comfortable deploying a Node.js and Postgres service
Not ideal for
- Individuals who only want a simple personal chatbot, since the setup targets organization deployments
- People without cloud deployment experience, because initialization runs through infrastructure, sign-in and connector steps
- Teams needing a fully managed product with no infrastructure work, unless they use the third-party hosted version mentioned in the README
How to get it
- 1
Install with a package manager
You'll need: Node.js 24.15+ (from the README badge) · npm · A cloud account (Fly or AWS) for deployment · Postgres for durable sessions
Initialize an organization deployment repository
npm exec --yes --package=@yc-software/qm@latest -- qm init . --org <slug> --target <fly-or-aws>Install dependencies in the deployment repository
npm install - 2
Run it
- Create an organization-owned deployment repository and run the init command above
- Follow the generated deployment skill through infrastructure, web sign-in and connector credentials
- Optionally add Slack access during setup
- Deploy and run the live verification step the initializer walks through
- 3
Get the most out of it
- Set DATABASE_URL and SESSION_STORE=postgres so sessions survive restarts
- Choose a security posture (Strict, Auto or Dangerous) that matches your risk tolerance
- Enable content screening via securityScreen.mode if you handle external content
- Use shared skills and admin-gated promotion to share useful skills across the org
- Add crons, watches and webhooks for background work while you are away
Downloads
Strengths
- Runs in your own cloud with your own models and keys, so it is not tied to one vendor
- Scoped memory, files and permissions per person and per room
- Clear security postures and an audit trail described in the README
Trade-offs
- Setup involves cloud infrastructure, Postgres and deployment decisions, so it is not a quick install
- Sessions live in memory unless you configure DATABASE_URL and SESSION_STORE=postgres
- Security defaults and modes need careful review before use with sensitive data
How it compares
| Project | Stars | Type | Level | Updated | Best for |
|---|---|---|---|---|---|
| qm (this) | 15.6k | AI model / agent | advanced | 4 days ago | Startups that want a shared AI agent for the whole team in Slack |
| deepagents | 30.9k | Library | intermediate | 4 days ago | Developers building AI agents that run multi-step tasks such as research and writing |
| openai-agents-python | 30.5k | Library | intermediate | 5 days ago | Python developers building multi-step AI workflows |
| pydantic-ai | 21k | Library | intermediate | 4 days ago | Python developers building LLM-powered features that need structured, typed output |
| openclaw | 394k | App | intermediate | 4 days ago | People who want an AI assistant that runs on their own machine rather than a hosted service |
| CowAgent | 47.8k | App | intermediate | 4 days ago | People who want a self-hosted personal assistant reachable from their everyday chat apps |
Stats synced from GitHub. The assessment above was drafted with AI from the project's README on 2026-10-29 and may miss recent changes — check the README before relying on a command. MIT licence: free to use, modify and distribute, including commercially, as long as you keep the copyright notice.