QM

QM is a self-hosted AI agent platform for startups that want employees to work with agents individually and in shared channels or projects. Each person and room has its own memory, files, permissions and persistent workspace.

• Personal and shared work: Collaborate with agents in Slack or a web app while keeping resources scoped unless they're shared, with identity and configuration carried between both interfaces.
• Search and background tasks: Search notes, email, documents, databases and the web, or schedule inbox labeling, reply drafts and project follow-ups.
• Apps and coding: Have agents build internal web apps for selected people, work in existing repositories, run tests, open pull requests and monitor CI.
• Organization controls: Choose available models and sharing rules, share reusable skills, and require human approval for tool calls; optional content screening can flag external material.

QM runs in your organization's own cloud account and is open source under MIT. It supports Pi, OpenCode, Codex and Claude Code, with your choice of models and keys. Agents use the credentials and permissions of the person they work for, and QM audits their actions.

Postgres can retain user data and session history. Without that persistence, sessions live in process memory and disappear on restart.