What it does
QM is an open-source, MIT-licensed agent harness built for shared work. Instead of one personal assistant, each employee gets an isolated workspace, and the agent can also join channels, group messages and projects. Every person and room has its own scoped memory, files, keychain view, permissions, scheduled jobs, web apps and durable sandbox. A headless core handles the API, identity, policy and scheduling, with Postgres holding sessions, memory and the queue. Pi and Claude Code can both drive the agent loop, and Pi runs models from many providers, so a deployment is not tied to one vendor. Slack and the web app share the same identity and configuration. An organization picks a security posture: strict approval for every tool call, an auto default that blocks private-network access, or a dangerous mode without approval gates. Sharing is isolated by default, and content screening can be off, observe-only or enforcing. You deploy into your own cloud account with the qm CLI, with Fly and AWS as the named targets.
Capabilities
- Personal and shared scopes, each with its own memory, files and permissions
- Slack and web app sharing one identity and configuration
- Replaceable harnesses and models, including Pi and Claude Code
- Strict, auto and dangerous security postures with audited actions
- Shared skills, crons, watches and inbound webhooks
- Self-hosted deployment through the qm CLI
When to use it
- Startups that want one shared agent across Slack channels, projects and individual workspaces
- Teams that need to run the agent in their own cloud with their own models and keys
- Organizations that want org-level control over models, approvals and sharing
