Agentcode

How it works

Autonomous coding agent, from task to pull request

Agentcode is an autonomous coding agent that works like an AI software engineer on your team. You describe the work; it plans, edits, tests, and opens a reviewable pull request. You review and merge.

In short

An autonomous coding agent completes a coding task end to end instead of suggesting one line at a time. Agentcode reads your codebase, writes a short plan, edits the files it needs, runs your real test suite, and opens a pull request with the plan, the diff, and a green test run. You review and merge; the agent never merges on its own.

  1. 1

    Describe the work

    Write a task in plain language: a feature, a bug, a refactor. Drop in a ticket or a one-line description. The agent reads the relevant parts of your codebase to understand the context.

  2. 2

    The agent plans it

    Agentcode writes a short numbered plan of how it will make the change, which files it will touch, and how it will verify the result, so you can see the approach before any code is written.

  3. 3

    It edits and runs the tests

    The agent edits the files across your codebase and runs your real test suite. It iterates until the tests pass; if it cannot get them green, it tells you what blocked it instead of shipping.

  4. 4

    It opens a pull request

    You get a pull request on a branch named for the task, with the plan, the diff, and a green test run. Review it, request changes, and merge when you are ready. The agent never merges itself.

See it run

Watch a task become a pull request

Pick a task and watch the agent plan, edit, test, and open a PR.

Agent Run

Pick a task

Plan

  • planning

Files changed

Test run

0 failed

Pull request

Open

You review and merge. Agentcode never merges on its own.

What it can do

Every part of the loop, broken out

Ready to clear the backlog?

Connect a repo and let the agent open its first PR.