Production deployment checklist

Run through this checklist after the single-node quickstart and before routing real repositories or user work through a self-hosted deployment. The goal is to verify that credentials, persistence, workers, observability, and rollback paths are ready before agents touch production code.

Core services

• API and frontend are reachable over HTTPS.
• Postgres migrations have run successfully.
• Redis is reachable by API and worker processes when configured.
• Workers can claim jobs and report health.
• For single-node, MODE=all is running and logs include worker started.
• Logs are shipped to the configured log backend.
• Durable session snapshots and uploads are backed by either S3-compatible storage or a backed-up single-node SINGLE_NODE_DATA_DIR mount.

GitHub

• GitHub App credentials are present.
• Webhook secret matches GitHub.
• Webhook delivery succeeds.
• Test repository import succeeds.
• A test PR can be opened from a session.

Workers and sandboxes

• Sandbox runtime health check passes.
• Single-node production uses SANDBOX_RUNTIME=runsc and SANDBOX_REQUIRE_GVISOR=true.
• Production worker-capable processes have SESSION_EXECUTOR_IMAGE configured.
• Session executors can reach Postgres and any required same-host storage.
• Worker concurrency is configured for host size.
• Disk limits and cleanup are enabled.
• Preview routing works for a simple app.
• Worker logs include enough context for debugging.

Rollback

Have a rollback plan for:

• API/frontend release.
• Worker release.
• Database migration.
• Secret rotation.
• Provider outage.