QM
Deploy QM, Y Combinator's multiplayer agent harness, on E2B. Every person and every room gets its own sandbox as the computer the agent works on.
- E2B
- Prebuilt QM templates
- qm-core, qm-sandbox
- 8 vCPU · 8 GiB memory
Get started
Paste the prompt into your coding agent, or follow the steps yourself.
Paste this into your coding agent
Point this QM deployment at E2B sandboxes.
- 1.In qm.config.jsonc, set "sandbox": { "backend": "e2b" } and E2B_TEMPLATE_ID to a pinned qm-sandbox version in env.core.
- 2.QM needs the E2B_API_KEY secret. If it isn't set, ask me for my API key from https://console.e2b.dev/?tab=keys and set it with qm secrets set. Never hardcode it.
- 3.Follow the QM guide at https://docs.e2b.dev/agents/qm. If my plan caps sandboxes at one hour, also set E2B_MAX_LIFETIME_SEC=3600 in env.core.
- 4.Send one message to QM and show me the output that came from its sandbox.
Or set it up yourself
1. Create the setup sandbox
2. Or point an existing QM at E2B
3. Look inside the agent computer
What runs in the sandbox
QM core keeps the org's state and gives every scope its own sandbox to work in.
- Each qm-sandbox ships Claude Code, Codex CLI, GitHub CLI, AWS CLI, Node.js 24 and Python 3.11.
- Core creates one sandbox per person and per room, and finds it again by name on the next turn.
- Conversations, memory and permissions stay in core. Recovery snapshots carry a scope's files across pauses and lost sandboxes.
- Sandboxes pause instead of stopping at their timeout and resume on the next turn.
More coding agents
Other coding agents that run in E2B sandboxes.
- Claude CodeRun Claude Code in its own sandbox with a real filesystem, terminal, and git, so it can work unattended.
- OpenAI CodexRun OpenAI's Codex CLI headless in a sandbox and collect the diff when it's done.
- OpenClawHost an always-on OpenClaw agent in a sandbox and chat with it from its web UI or Telegram.
