This is an early release preview. You may encounter bugs.
Qm logo
Unclaimed

Harness coding

Qm

Y Combinator's multiplayer agent harness multiplayer AI agent harness.

Built by Y Combinator

B 76/100 GitHub score ? This grade is derived from GitHub signals, not user votes. Open for the full breakdown.
Discussion

01 / Summary

About Qm.

QM is a multiplayer agent harness for companies, used in Slack and on the web and deployed in the organization's own cloud with its own models and keys. Each employee gets an isolated personal workspace, and people also work with the agent together in channels, group messages, and projects. Every person and every room has its own scoped memory, files, keychain view, permissions, crons, web apps, and durable sandbox.

Every turn runs through a headless core that handles the API, identity, policy, and scheduling, and hands the agent loop to an interchangeable harness: Pi, OpenCode, Codex, or Claude Code. Postgres holds sessions, memory, and the work queue. The agent has a small fixed tool surface, including an execute tool that runs commands in the scope's sandbox, a durable computer where installed tools stay installed. Harness, session store, sandbox, and memory each sit behind an interface, and memory can route to external providers by scope.

Like a local coding agent, QM acts as the person it works for, with their credentials and permissions, and audits everything it does. An organization picks a security posture (Strict, where every tool call waits for approval, Auto, or Dangerous) that narrower scopes can only tighten, plus a sharing posture that keeps resources isolated by default. Optional content screening classifies external content and can quarantine flagged items. Org-specific configuration, tools, skills, and sandbox images live in a deployment directory that the qm CLI validates and deploys to Fly or AWS.

  • Personal and shared scopes: each person customizes their own agent and also works with it in Slack channels, group messages, and projects
  • Slack and web: the same identity and configuration carry between Slack and the web app
  • Interchangeable harnesses: Pi, OpenCode, Codex, and Claude Code drive the same core, and admins choose which harnesses and models are available
  • Durable sandboxes: each scope gets its own sandbox where installed tools and files persist
  • Security postures: Strict, Auto, or Dangerous at org level, with a command policy that blocks destructive operations in every posture
  • Content screening: optional classification of external content that observes or quarantines flagged items
  • Web apps: build internal apps and publish them to the right people
  • Shared skills: scope-owned skills shared by grant, with admin-gated promotion to the whole org and skill packs imported from Git repositories
  • Background work: crons, watches, and inbound webhooks run while you are away
  • Deployment directory: org config, custom tools, skills, and sandbox image validated and deployed by the qm CLI

02 / Discussion CREDIBILITY-GATED

Discussion

Reading is open to everyone. Posting and voting need a verified identity or a GitHub grade of B or higher.

  • No discussions yet.

04 / Build

Build with Qm.

Browse the catalog for frameworks, tools, and harnesses, each scored on real GitHub credibility.

Get Qm →

Browse the catalog