# qm

A multiplayer AI agent harness for a company's Slack and web workspace, giving each employee and each room its own scoped memory, files, permissions and sandbox while sharing one underlying agent core across any supported model or harness.

qm is a multiplayer agent harness built for a whole company rather than one person, working in Slack and on the web. Each employee gets an isolated workspace and can also collaborate with the agent in channels, group messages and projects, with the same identity and configuration carrying over between Slack and the web app.

Each person and each room has its own scoped memory, files, keychain, permissions, scheduled jobs and durable sandbox. Admins set org-level configuration, security and sharing policies, and choose which agent harnesses and models are available; Pi, OpenCode, Codex and Claude Code can all drive the same core, so a deployment is not tied to one vendor. Skills are scope-owned and can be shared by grant, promoted org-wide by an admin, or imported from a git repository, and background crons, watches and inbound webhooks keep running while a user is away.

Typical uses include searching internal notes, email, documents, databases and the web together, building and publishing internal web apps, triaging an inbox with drafted replies in a learned writing voice, and working in an existing code repository to run tests, open pull requests and monitor CI. It is written in TypeScript and released under the MIT license.

Website: https://github.com/yc-software/qm