Treq is your AI Code Review Manager, accelerating AI-assisted software development while maintaining high quality code. Treq is an open-source Graphite alternative with a focus on paralleized development using AI agents.
Treq was used to build Treq.
Pre-requisites:
- Git
- Jujitsu You don't need to know how to use it
Download the latest release here.
Inspect and iterate on each change for a human-in-the-loop agentic workflow.
- Review the code diffs just like a familiar Github PR, annotate and comment on code, and then send it to an agent for changes.
- Spotted an issue when browsing files? Send it to an agent for adjustments in the background.
- Got a specific commit that you want the agent to work on, review it and send your comments directly to the agent. Did I mention that we have commit management tooling as well?
Coding agents work in isolated copies of the codebase, ensuring changes are independent from each other while keeping your current repository clean for planning.
-
Workspaces are isolated but and automatically rebased, meaning code never goes stale.
-
Got a code conflict? Agents can automatically resolve your conflicts.
-
Stack workspaces and build features incrementally over smaller bite-sized branches. All stacked workspaces automatically rebases on top of each other, even on top of your uncommitted changes.
# to profile benchmark code
samply record cargo bench --bench sync_workspacesLicensed under the Apache License, Version 2.0