Set up Agent Studio and review your first task
Start the Studio with Docker Compose, connect an execution host and a repository, then review one small change.
Setup documentation checked on 25 September 2026. The commands below follow the product README at that date; this website update did not include a fresh installation test.
1. Prepare the environment
Have Git, Docker with Compose, and at least 8 GB of free disk space available. The Compose path uses pinned release images and needs no host .NET or Node.js installation.
- Choose a disposable repository or a clean checkout with a recoverable Git baseline.
- Install and authenticate the coding-agent CLI on the execution host. Verify it starts outside the Studio.
- Model access and provider charges are configured separately from Agent Studio.
2. Start the Studio and add execution capacity
Run the commands below, then open http://localhost:4011. This starts the Studio services. Coding tasks also require a configured execution host with repository access and an authenticated CLI.
For a Docker runner, follow section 5 of the setup guide: bootstrap it with scripts/compose-runner-bootstrap.sh, then start the runner profile with docker compose --profile runner up --wait. The guide covers credentials and persistence.
Add your target repository as a project and verify that its assigned host can access it. A visible board alone does not confirm that execution is ready.
git clone https://github.com/agent-orc/agent-studio.gitcd agent-studiodocker compose up --waitWhat runs where
The browser displays tasks and review decisions. The Task Server stores task state and run records. Agent Runners execute coding-agent CLIs on configured hosts in isolated Git worktrees.
System map
Central state. Execution anywhere.
Control stays central. Execution lives wherever you allow it.Agent Studio
See and decide in one browser surface.
HTTP client Task Server
Remember and coordinate every task, run and review decision.
authoritative state Agent Runner
Execute coding-agent CLIs and report from a host you control.
agent-runner3. Define one observable change
Create a task with one outcome, a clear file or feature boundary and a relevant check. For example: improve the project list’s empty-state message, preserve its behavior and provide a mobile screenshot.
4. Run, inspect and decide
Select the configured agent and host. Follow the live run state, then inspect the diff and recorded checks. Accept the result or request a specific correction. Your first completed trial is a reviewed task with an explicit decision.
Exception example: DEMO-9 has exhausted its reissue budget. Open findings and pending integration remain visible beside the decision controls.
If setup or execution fails
Keep the first concrete error, command and output. Check the same CLI command outside the Studio, the project path and the host’s repository permissions before changing the task.