Greptile reviews pull requests with repository context, custom rules and integrations for development teams.
Greptile is an AI review service for pull requests. It indexes a repository into a graph of files, functions and dependencies, then uses that context to examine changes beyond the immediate diff. The official workflow describes parallel review agents looking for potential issues, including logic problems and violations of team conventions. The result is review feedback for developers to investigate, not a guarantee that every defect or security issue has been found.
Teams can provide custom rules and repository-specific context. Greptile also describes learning from existing engineer comments so that reviews better reflect local standards. GitHub and GitLab are supported, while MCP and coding-agent connections help pass issue context into tools used to make fixes. The current site introduces TREX as an agent that writes and runs tests in a sandbox, but labels access as early access. That distinction should be preserved rather than assuming every account immediately includes unrestricted autonomous testing.
The free Starter plan is limited to one active developer with a monthly credit allowance. Pro is priced per seat, and standard reviews and TREX reviews consume different numbers of credits. Enterprise arrangements include additional controls and self-hosting options. Before connecting private repositories, review the selected deployment's access and data handling. Start with a pull request whose risks are already understood, evaluate useful findings and dismiss incorrect ones with clear feedback. Keep normal tests and human approval in place. An automated reviewer can broaden coverage and shorten a feedback loop, but its suggestions still need validation against the implementation, requirements and behavior of the complete system.
