How much of your code is written by agents?

Buildermark matches your coding agent logs with git commits to attribute every line — local and open source.
No workflow changes. No agent hooks.

What it does

A measurement layer for coding agents.

Agent attribution

Track the exact percentage of each commit that was written by an AI coding agent — no reliance on agent hooks, no manual tagging.

Conversation ratings

Rate agent conversations manually, or let the agent log its own critique with the /rate-buildermark skill.

Agent benchmarking

Compare how Claude, Codex, Gemini, and Cursor perform on your actual codebase.

Supported agents

Every agent you use.

  • Claude Code CLI
  • Claude Code Cloud
  • Codex CLI
  • Codex Cloud
  • Gemini CLI
  • Cursor

Missing one? File a request →

Local-first

Your data never leaves
your machine.

  • Zero analytics, zero telemetry
  • Runs entirely on localhost:55022
  • MIT licensed, audit-friendly
  • Only outbound request: checking for updates
How it works

Fully Automatic

  1. 01

    Auto-import conversations

    Chat history from Claude Code, Codex, Gemini, and friends is imported automatically. Add shared folder paths to pull from VMs and containers.

  2. 02

    Auto-import commits

    Your git history loads automatically. Nothing leaves your machine.

  3. 03

    Auto-match diffs

    A formatting-agnostic matcher aligns conversation diffs to commit diffs — robust against auto-formatters, reorderings, and line-ending noise.

Native notifications

Attribution in your notification center, the moment you commit.

See agent percentage for each commit immediately — no dashboard required. Works across macOS, Windows, and Linux.

Buildermark native notification
Desktop apps
  • macOS 15+
  • Windows 10+
  • Linux CLI
Browser extensions
  • Chrome
  • Firefox
  • Safari

For importing from cloud-based agents

Architecture

A local app container runs a Go server on localhost that serves the web UI. No cloud. No accounts. No sync.

Coming soon

Team Server.
Aggregate. Compare. Compete.

A self-hosted server that rolls Buildermark metrics up across your whole org. See who's shipping with agents, which projects adopt fastest, and how your dev process is changing — all on your own infrastructure.

Join the waitlist →