YC releases qm, a multiplayer agent harness for startups
Y Combinator's yc-software team open-sources qm, a multiplayer agent harness that lets startup teams collaborate with AI agents in Slack and on the web.
BY FOUNDERBUILT AI NEWS
Y Combinator's yc-software team has open-sourced qm, a multiplayer agent harness built for startups. Unlike personal-assistant style agents, qm gives each employee an isolated workspace with scoped memory, files, keychains, permissions, crons and durable sandboxes, while letting teams collaborate with the agent in Slack channels, group messages and shared projects.
The harness is deliberately vendor-neutral. Pi, OpenCode, Codex and Claude Code all drive the same core, so a deployment is never tied to one model provider. A headless TypeScript core on Node handles identity, policy and scheduling, with Postgres for durable state, while the web UI, admin panel and Slack integration plug in as optional components.
Security follows the local coding agent model: the agent acts as the person it works for, and everything it does is audited. Orgs pick a posture from Strict, where every tool call pauses for approval, to Auto, which screens external data with a classifier, to Dangerous with no screening. With org-level admin controls, shared skills and background crons, qm is aiming squarely at teams that want agents embedded in daily work without locking into a single vendor.