Overnight execution

Run AI coding agents overnight without granting a blank check.

A safety-first overnight supervisor for Codex, Claude Code, Cursor, Gemini CLI, OpenCode, Copilot CLI, and CI.

See a supervised run Browse integrations

Without a supervisor

The problem

An eight-hour prompt is not an operating model. Agents stall, tests fail, sessions expire, and permission dialogs wait forever. Broad auto-approval solves the waiting by creating a larger risk.

With Sleep Mode

The operating model

Sleep Mode turns the night into a bounded run: one goal, one deadline, a signed permission ceiling, durable liveness, and a morning brief backed by the actual event record.

How it works

From vague autonomy to bounded operations.

Arm the outcome

Define the repository, goal, cost, action, time, command, network, and approval boundaries before stepping away.

Keep the same sessions

The local daemon resumes identity-bound provider conversations instead of silently starting over.

Recover the routine

Policy-selected nudges and fixed test/build actions handle ordinary stalls while hard boundaries wait.

Wake up oriented

See verified progress, spend, interventions, failures, and the exact next human decision.

Safety evidence built into the run

Four-minute fail-safe watchdog One-minute signed heartbeats Hard deadline Workflow Evidence-linked signed brief