Skip to main content
Two catalog agents work against your repositories: Code Reviewer, which reviews pull requests as they change, and GitHub Agent, which finds where a characterised problem lives in the code. Neither writes to a repository beyond Code Reviewer’s own review comment.

Code Reviewer

On each push to a pull request, it reads the diff and reports what a maintainer needs before merging. Findings come back in three buckets, each naming a file, a line, and the concrete consequence rather than a style preference: It comments once per commit rather than amending one review, so the history of what was flagged and when stays readable on the pull request itself.

The verdict

The worst finding decides the verdict, and the agent does not choose it — the buckets are what it produces, and the verdict follows: Finding nothing is a valid and common result. The review says so rather than padding itself to look thorough. Where it stops. It never casts a GitHub approval — its verdict is reported in its own comment, and “Looks good” is a verdict rather than an approving review, so it cannot satisfy a branch-protection rule that requires one, and it never merges. Posting that comment is the only thing it writes: no pushes, no branches, no file changes.

GitHub Agent

Give it a characterised problem — an incident, an error message, a failing service — and it searches code for the error’s own text, opens every candidate to confirm it rather than trusting the search ranking, and checks what changed there just before the problem started. What comes back is a ranked shortlist of files with the evidence for each, and the durable code change that shortlist points to. Read access only.
It runs on concrete evidence. Given a vague problem statement with no error text, no service and no timeframe, there is nothing distinctive to search for — so pass it the output of an investigation rather than the original alert.
Before you deploy — model access, and a GitHub connection with read access to the repositories it searches. The connection is required and gates the agent’s only path to GitHub: without one it cannot search at all, and it fails closed and says so rather than returning an empty result. Where it stops. It never modifies a repository — no branch, no commit, no pull request. Opening the pull request that applies a fix is Remediation Executor’s job. It also does not tell you what to do right now: its recommendation is the durable code change that stops a problem recurring, which is a different question from how to restore service during an incident.

Next steps

Agent catalog

Every deployable agent, and what each one is for.

Integrations list

Connecting GitHub.

Remediation Executor

Turning a recommendation into a pull request, with approval.

Approvals

The decision boundary before anything is written.