You Asked, We Shipped: Native Grok Support and an Auto-Approve Switch
Two features went live today, and both exist because customers asked pointed questions this week. That's our favorite kind of roadmap.
Native Grok support — your xAI key, plugged straight in
CodeMouse now speaks to xAI directly. Add your xAI API key in Dashboard → API Keys, pick a model — Grok 4.7 (the flagship), Grok 4.3 (a 1-million-token context window, great for very large diffs), or Grok Build 0.1 (the budget pick, tuned for code) — and Grok becomes a full voice in your review consensus, right alongside Claude, GPT, Gemini, or anything you run through OpenRouter.
Grok gets the full machinery: its diff window is sized to the model's real context, oversized pull requests are reviewed in multiple passes and merged into one review, and findings are de-duplicated against your other models. Costs are computed from actual token usage at xAI's published rates and recorded on every review — billed to your key, with no markup, like every provider on CodeMouse.
(If you were already running Grok through OpenRouter: that keeps working. Native support just skips the middleman — handy if you have xAI credits or want one less account in the loop. One note: xAI has no batch API yet, so like Gemini it sits out economy mode.)
Fun fact: someone asked "any plans to add Grok?" on our AppSumo listing yesterday morning. It shipped the same afternoon. xAI's API is OpenAI-compatible, and sometimes the stars align.
The auto-approve switch
CodeMouse has always been willing to approve a pull request — but only under strict conditions: every model you've configured independently recommends approval, there are zero findings, every CI check is green, and the review covered the full change set. Plenty of teams love that; a clean PR merges without waiting on a human.
Some teams, though, have a hard rule: software never approves software. Compliance requirements, review policies, or simple taste. Until today those teams had no way to use CodeMouse without also accepting its approvals.
Now there's a switch. Turn Auto-approve off in Dashboard → API Keys and CodeMouse keeps reviewing every pull request automatically — findings, suggested fixes, thread resolutions, all of it — but posts everything as comments only and never submits an approving review. The final "looks good" stays with a person, forever, org-wide.
Also this week, in the honesty department
While we were at it, we rewrote our privacy policy's code-access section to describe exactly what happens during a review (temporary isolated clone, deleted the moment the review ends, nothing retained), and expanded the docs FAQ with the questions buyers actually ask: what CodeMouse finds, when reviews run, what precisely is sent to the AI providers, and whether your code trains AI models (it doesn't — your keys, and the major providers' API terms don't train on API data by default).
Keep the questions coming. This week they turned into two features, a clearer privacy policy, and a better FAQ — the suggestion box is [email protected].
FAQ
Which Grok models does CodeMouse support natively?
Grok 4.7, Grok 4.3, and Grok Build 0.1, selected in Dashboard → API Keys after adding your xAI key. Other Grok variants remain available through OpenRouter.
Does turning off auto-approve change what the reviews contain?
No. Reviews are identical — same findings, same suggested fixes, same thread resolutions. The only difference is that CodeMouse never submits an approving review, so merging always waits for a human approval.
Do I pay anything extra for Grok reviews?
Nothing to us. Inference is billed by xAI directly to your key at their standard rates; CodeMouse records the per-review cost in your dashboard and adds no markup.
Is Grok's context really 1 million tokens?
Grok 4.3's is, and CodeMouse takes advantage of it: the diff window scales with each model's context, so a large PR that needs multiple review passes on a smaller model can often fit in one pass on Grok 4.3.