The role has changed

Agents write the code.You make the calls.

You say what to build and judge whether it's right. Most people just lack the infrastructure to run that loop unattended.

$curl -fsSL reeve.run/install.sh | bash
Runs on your machine, with your API keys.Use Claude today, swap to Codex tomorrow. Pipeline stays the same.Agent crashes, Reeve picks up from the checkpoint.
Watching studio/web-app
[12:04] ● REV-42 "Add OAuth provider support"
[12:04] ⠋ REV-42 Running claude…
[12:07] ↗ REV-42 PR → studio/web-app#89
[12:08] ✖ REV-42 review failed
[12:08] ↻ REV-42 Retrying (1/2)
[12:10] ✔ REV-42 review passed
[12:11] ✔ REV-42 Done (merged)
The new loop

Reeve connects your issue tracker to your agent in the background. You write tickets; it isolates the work, runs the agent, and delivers a PR. It recovers from crashes and never touches your main branch.

Any SourceBring your own issue tracker
Any AgentToday's strongest, tomorrow's too
IsolatedEvery task runs in its own environment
ReeveTakes the job, watches it, retries on failure
Your CallReeve delivers PRs, merging is yours
The bet

Between complex governance and pure execution, Reeve focuses on the latter.

Most platforms are building thick governance layers on top of agents. Reeve stays deliberately thin: it isolates a task, runs the agent, and reports back. There is no intelligence in the orchestration layer to tune or outgrow, so when agents improve, your pipeline improves with them.

Agents get stronger every month. A thin layer lets progress pass through. A thick one becomes a ceiling.

42 seconds

A quick look at how Reeve runs the loop.