Stop repeating environment setup
Reuse dependency definitions across tasks. Spend less time explaining your environment from scratch.
Building got easier.
But how does it all work?
Install with npm. Connect your coding agent. Review code changes, test results, and application captures together for each worktree.
A /redpact init request connects dependency definitions for Web, API, Event Server, and Worker, sharing one Database. These are not running environments.
Build quickly. Understand what works.
Connect the evidence you need, from setup to verification and review.
Connected to the agents you use
Keep using your coding agent. Connect tests and reviews through MCP.
An experience built for AI pairing
Review AI changes by worktree. Inspect diffs alongside test results and decide what comes next.
Containers, from setup to cleanup
Prepare containers for each test run. Clean up after completion, failure, or cancellation. Manage the container lifecycle in one place.
Before you get started.
Explore the docsWhat is Red Pact?
Red Pact is a local testing and review tool for changes made with a coding agent. Inspect test intent, execution results, screenshots, and code diffs for each task in one place to decide whether to accept the change.
Do I need to switch coding agents?
Connect Red Pact to your existing MCP-capable agent. MCP UI cards appear in hosts that support MCP Apps. Use the Red Pact viewer for detailed execution history and review.
How do I get started, and what do I need?
Install with npm or pnpm on the download page, connect your agent to the same Red Pact instance, and ask it to set up your project. Managed tests require Docker Compose and execution and test settings for your project. Terminal installation requires Node.js 24 or newer.
Which tests can I run?
Red Pact supports Unit commands, Integration tests, and Playwright functional tests and screen captures. The MCP run_tests tool runs Integration tests. Start Unit and Playwright runs from their respective execution screens, and review captures separately from functional test results.
Do I need to set up the environment for every worktree?
Reuse project dependency definitions and select modes for the task. New dependencies or test requirements still need configuration. Each run prepares and cleans up its managed temporary environment, but shared databases and external services need separate test data isolation.
Can I accept a change just because the tests pass?
Passing results provide evidence within the scope of the tests executed in the selected environment. A mock payment API does not verify the real provider. Review the executed source, results, screenshots, diffs, and unverified scope together. If the code changes again, it needs a new execution result.
Are this page and the demo real execution records?
This page and the app demo explain the workflow. They do not connect to your projects or perform real tests or Git operations. View execution records for your own projects in your installed Red Pact instance.
Redpact is open source.Join as an early contributor
Bug reports, ideas, and code — all contributions are welcome.
Contribute on GitHub