Sweep
Open-source AI junior dev that turns GitHub issues into tested pull requests.
About
Sweep is an open-source AI agent (MIT/Apache) that works like a junior developer inside your GitHub workflow. You write a bug report or feature request as a GitHub issue with a Sweep: prefix, and it searches your codebase with vector search and AST analysis, plans the implementation, writes the code and unit tests, and submits a pull request. It respects your linters and formatters, reads reviewer comments on its PRs, and pushes fixes in response. Behavior is tuned through a sweep.yaml config file.
The honest picture: Sweep shines on small, well-scoped tasks and chores like typehints or small refactors. Complex architectural changes and ambiguous requirements still need a human engineer, and as with every autonomous code writer, every PR deserves a real review before merging.
What it can do
- Turn GitHub issues directly into pull requests
- Vector search plus AST-based codebase understanding
- Runs unit tests and autoformatters to validate changes
- Responds to review comments with new commits
- Configurable behavior via sweep.yaml
Setup
Install the Sweep GitHub App on your repositories. Open an issue starting with Sweep: and it picks up the work automatically.
Prerequisites
A GitHub repository with issues enabled. Python projects get the best results.
Use cases
Works with
Trust score — 7/10
Scored 7/10 (metrics tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 48; Trending score 68/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.