Skip to main content
Indent can pick up a coding task, work through it in a real repository, and carry it all the way to a merged pull request. The session stays connected to the PR it creates, so Indent can keep iterating — on your feedback, review comments, and CI failures — until the work is done. Best with: Environment Setup. Add AGENTS.md, Skills, related repositories, or Databases when the task needs more context.

​
From prompt to pull request

Describe a bug, feature, or task in chat. Indent explores the codebase — reading files, searching for references, running commands — then plans the change, edits the code, runs linters and tests, and opens a pull request. You’ll see each step as it happens: which files Indent read, what it edited, what commands it ran, and why.

​
The feedback loop

Once the pull request is open, CI status, review comments, and @indent mentions on the PR are sent back to the session automatically — so you don’t have to relay feedback manually.
  • Review comments — when a teammate comments @indent on the PR, the message routes back to the session that created it. Indent picks it up with full context and can push a follow-up commit without anyone having to re-explain the task.
  • CI failures — when a check fails, Indent can read the logs, figure out whether the failure is related to the PR or an infrastructure flake, and push a fix or flag it for you.
  • Your follow-ups — you can send messages in the session to refine the approach, ask for changes, or redirect the work entirely. If the session has suspended, sending a new message picks up where it left off.
This means you can steer Indent the same way you’d work with a teammate over the course of a PR — not as a one-shot prompt, but as an ongoing conversation until the code is ready to merge.

​
Bugs, features, and everything in between

Indent operates in a full sandbox environment with access to your repository, a shell, a browser, and any tools configured in your environment setup. That means it can handle:
  • Bug fixes — reproduce the issue, trace the root cause, fix it, and verify.
  • Features — scaffold new functionality, wire it into existing code, and run tests.
  • Infrastructure and config — edit CI pipelines, Dockerfiles, deployment configs.
  • Assets and non-code files — SVGs, config files, migration scripts — anything in the repo.