What is OSU?
OSU sits between your AI coding agents and GitHub. When you run more than one agent (Claude Code, Cursor, Replit, Lovable), or anything else pointed at your repo they start stepping on each other: hitting the same rate limits, pushing conflicting branches, and merging out of order with no record of who did what. OSU is the coordination layer that stops that from happening, without asking you to change how you already work with any of these tools.
It's not another agent, and it doesn't replace GitHub. It's a GitHub App that queues, locks, and audits every agent action before it touches your repo.
How it's used
Once connected, OSU works quietly in the background. Every job your agents submit, a commit, a branch, a merge, goes through OSU first. It gets queued against the right repo, checked against your remaining GitHub API budget, run in an isolated sandbox, and rebased cleanly before it ever lands. If two agents try to touch the same repo at once, OSU sequences them instead of letting them collide. If a merge would conflict, you see it flagged, not silently broken.
Everything that happens is written to an audit log you can't edit or delete, so at any point you can see exactly which agent did what, when, and whether it succeeded.
You don't manage queues, locks, or sandboxes directly, they're automatic. What you interact with is the dashboard, where you can watch agent activity live and review history.
How to use it
- Sign in: Sign in through your GitHub account first. This is what your repo access and audit history are tied to.
- Connect a repo (see below).
- Point your agents at OSU: No change to how you prompt or run them, just route their GitHub actions through OSU instead of straight to GitHub.
- Watch it work: Open the dashboard to see live activity across every connected repo: what's queued, what's running, what merged, and what needs your attention.
How to connect a repository
- From your OSU account, choose Connect a repository.
- Install the OSU GitHub App on the repository, or the whole organization, if you want every repo covered.
- Grant access: OSU reads your existing GitHub collaborator list, so permissions on OSU always match permissions on GitHub. Nobody gets access here they don't already have there.
- That's it. The repo now has a dedicated queue, its own rate-limit budget, and its own audit log from the moment it's connected.
If you remove someone's access on GitHub, it's mirrored on OSU automatically, you don't manage two separate permission lists.
How agents are integrated
OSU works with any agent that can be pointed at a GitHub App instead of a personal access token, including Claude Code, Cursor, Replit, and Lovable. There's no plugin to write and no agent-specific setup: from the agent's point of view, it's just talking to GitHub through a different door.
What changes is what happens after the agent submits its work:
- It's queued, not applied instantly, so a second agent working the same repo can't collide with it.
- It's rate-limit aware: OSU tracks your GitHub API budget across every agent using the repo, not just one.
- It runs isolated: Each agent's job executes in its own sandboxed environment, so one agent's in-progress work can't bleed into another's.
- It's rebased, not just merged: History stays linear, and conflicts are caught before they hit your main branch.
- It's logged: Every action any agent takes is recorded permanently, so multi-agent work stays traceable even when three tools touched the same repo in one afternoon.
Because OSU treats every agent the same way, you can mix and match tools freely, run Claude Code on one feature and Cursor on another, in the same repo, at the same time, without coordinating them yourself.