QM — multiplayer agent harness
Scopes, headless core, harnesses, security postures, and the deployment CLI mental model for QM.
Questions
Plain-language prompts grounded in this codebase. Open any question for a full answer page.
- ClozeL1
QM is a multiplayer agent harness for work, available in Slack and on the web.
- ClozeL1
Each person and each room has its own scoped memory, files, keychain view, permissions, crons, web apps, and durable sandbox.
- ClozeL1
Every turn runs through a headless core that owns identity, policy, scheduling, and the agent loop.
- ClozeL1
Pi, OpenCode, Codex, and Claude Code can all drive the same QM core.
- ClozeL1
The QM core runs TypeScript on Node and uses Fastify for HTTP.
- ClozeL1
The default security posture is Auto, which screens provenance-labelled external data before it reaches the model.
- ClozeL1
Org-specific config, sandbox image, and infrastructure live in a deployment directory that the qm CLI validates and deploys.
- ClozeL1
Headless core runtime code lives primarily under the src/ directory.
- ClozeL1
Contributions are taken as human-written text in adrs/, not as code pull requests.
- Multiple choiceL2
How are shared skills owned and promoted in QM?