Building Dashboards
Git AI tracks AI code across the entire SDLC, providing a clear measure of the effectiveness, adoption, and long-term durability of the code written by each Agent.
When you are ready to collect metrics from multiple engineers and across repositories, you can export OpenTelemetry events from your SCM (GitHub, GitLab, etc) to any OTEL-compatible dashboard.
For a PoC or a rollout to a limited number of repositories, you can use a CI job configured per-repo to export these metrics. If you want to collect metrics from all of your repositories, the Self-hosted SCM App + Webhook is recommended.
| Stage | Scope | Approach |
|---|---|---|
| PoC / Limited Rollout | 1-5 repos | CI job configured per-repo |
| Wall-to-Wall Rollout | or thousands of repos | A Self-hosted SCM App that recieves webhooks from every repository |
Both approaches processes keep your code secure, processing all code and metadata on compute you manage.
Metrics Overview
For reporting and analytics, Git AI aggregates the metadata it stores in its Git Notes and exports metrics to any OpenTelemetry (OTEL) compatible dashboard.
These OTEL events Git AI collects can help you answer questions like:
- What % of each PR is AI-generated?
- What % of all the code in the repository is AI-generated? Trendlines by date, tool, model, contributor, etc.
- Compare ROI of each AI-agent / LLM.
- What is the half-life of AI-generated code? (eg. Is it durable? Do we just revert it?)
- How many of our [Copilot | Cursor | Claude Code]-Licenses are being used?
- Which AI-agents / Models produce the most accepted AI-generated code? Locally? In Code Reviews?
- Which AI-agents / Models waste tokens (eg. code reverted, or changed during Code Review)
- Adoption of AI tools by contributor / team. Trendlines by date, tool, model, contributor, etc.
- Token Usage and Total AI cost by PR
All of that data can be grouped by:
- Your entire Organization
- A specific set of SCM Organizations and/or Repositories
- Pull Request
- Model (eg. Claude 3.5 Sonnet, GPT-4o, etc)
- Tool (eg. Copilot, Cursor, Gemini, Claude Code, etc)
- Individual Contributor / Team
- Date Range (identify trends eg. Last 7 days, Last 30 days, Last 90 days, etc)
Get Started
Run the metrics exporter as a CI job for PoCs and limited rollouts (1-5 repos)
For rollouts to 10s, 100s or 1000s of repositories, the Self-hosted exporter is recommended. Deploy a Helm chart, and add a webhook to your SCM.
Self-hosted Exporter →