Docs
Environments you can version, fork, and grade.
Start with the quickstart: a deployed, forked, and graded environment on your own machine in about five minutes. Everything on these pages is copy-paste runnable.
Tutorials
Learn by doing. Start here.
- QuickstartDeploy an environment, fork it, and grade it on your own machine in about five minutes.
How-to guides
Get a specific thing done.
- Connect a targetAdd somewhere for environments to run — your own machines, a cloud account denv provisions in, or a hosted sandbox provider.
- Deploy from GitHub on pushConnect a repository so every push to the default branch deploys alpha, with a check run that links to the phase log.
- Run a Harbor dataset on denvUse denv as the Harbor environment backend so every trial gets a forked, protected environment.
- Use denv as an Inspect sandboxGive every Inspect sample a forked denv environment with sandbox="denv".
- Run an OpenEnv environment on denvUse DenvProvider to run an OpenEnv image as a denv sandbox and fork it mid-episode.
- Grade an episode with a separate verifierDeclare a verifier in the compose file and run denv verify so the grader never shares a process tree with the agent.
- Snapshot and forkCapture a running environment as a version, fork it per rollout, and roll back.
Reference
Every command, route, and field.
- denv CLIEvery denv command, subcommand, and flag, with the configuration the CLI reads.
- denvd sidecarThe sidecar binary that runs on a host: verbs, flags, what it does on the machine, the install script, and the systemd unit.
- REST APIEvery route of the denv control plane, with auth, request and response shapes, and a curl example.
- Compose file and x-denvThe compose subset denv accepts, what it refuses and why, resource caps, the x-denv block, and the verifier contract.
- Python packageThe denv Python client and its Harbor, Inspect, and OpenEnv adapters, method by method.
- Control plane ↔ sidecar protocolThe frames exchanged over a host's stream, every command kind with its payload and result, idempotency, and heartbeats.
- Deployments and phasesEvery deployment kind, its phases in order, what each phase records as evidence, and how failure is handled.
- TargetsWhere environments run — sidecar hosts and hosted sandbox providers — with the capability table, placement rules, and join tokens.
- Control plane configurationEvery flag and environment variable of denv-cp, the GitHub integrations it needs, and its CORS behaviour.
- ProtectionsThe five protections an rl-environment gets by default — what enforces each, where the evidence is, and the exact strings you will see.
Explanation
Why it works the way it does.
- ConceptsProjects, environments, deployments and versions, and why an environment is a system rather than a container.
- How snapshot and verify workWhat happens when denv captures a running system, and why grading happens on a fork.
- Data handlingWhat denv stores, what it never sees, and where your environments' data lives.