1
Install Greptile for the GitHub or GitLab repositories you want reviewed and allow the initial codebase index to complete. When a pull request is opened, review Greptile's comments alongside CI results and human feedback. Use repository rules to teach it project-specific expectations, and mark unhelpful findings so future reviews better reflect the team's standards. When sending a finding to another coding agent for a fix, review the resulting patch rather than treating the automated handoff as an approval.
2
For a first evaluation, keep the scope small enough that you can compare the AI-assisted result with a known baseline. Record the configuration, model or workflow choices that produced the result so successful tests can be reproduced. If the product connects to external systems or private data, grant the minimum permissions required and review the provider's current security and data-handling documentation before expanding access.