GitVelocity is a new tool that uses AI to read every merged pull request in a GitHub repo and assign a productivity score based on what was actually built. It's free to use if you bring your own API key.
The pitch: instead of measuring engineering output by lines of code, story points, or manager estimates, let an AI model analyze the substance of each PR across six complexity dimensions and produce a single velocity number. Individual engineers get personal dashboards tracking their output over time. Team leads get comparative views across people and time periods.
The "bring your own API key" model means GitVelocity itself is free, but you're paying your LLM provider (likely OpenAI or Anthropic) for the inference costs of analyzing your PRs. For a small team, that's probably a few dollars a month. For a 300-person engineering org, the API bill could add up.
The obvious question with any tool like this: does reducing engineering work to a single number create perverse incentives? Engineers optimizing for velocity scores might favor many small PRs over fewer, more thoughtful ones. GitVelocity claims it analyzes complexity rather than volume, but the specifics of how it weights different contribution types aren't fully transparent.
GitHub integration is the only supported connection right now. If your team uses GitLab or Bitbucket, you're out of luck for the moment. The tool targets teams from 3 to 300+ people, positioning itself as useful for both individual growth tracking and team-level management.
For engineering managers who want data beyond gut feel but don't want to build internal dashboards, it's worth a look. Just be thoughtful about how you use the scores with your team.