Caroline Bishop
Jan 29, 2026 21:35
Claude Code now tracks GitHub PRs and commits with AI assistance. Anthropic reports 67% increase in PRs merged per engineer internally.
Anthropic released contribution metrics for Claude Code on Thursday, giving engineering teams a way to measure how much the AI coding assistant actually impacts their output. The feature, now in public beta, integrates directly with GitHub to track pull requests and commits made with Claude Code’s help.
The numbers from Anthropic’s internal usage tell an interesting story. Since ramping up Claude Code adoption, the company says its engineers now merge 67% more PRs per day. Even more striking: 70-90% of code across teams is now written with Claude Code assistance.
What the metrics actually show
The GitHub integration surfaces three main data points: PRs merged (split between assisted and unassisted), lines of code committed with and without Claude Code, and per-user contribution patterns across teams.
Anthropic is being careful about attribution here. The system only counts code as “assisted” when there’s high confidence Claude Code was involved, matching session activity with GitHub commits. Workspace admins can access everything through the existing Claude Code analytics dashboard—no extra tooling required.
The productivity measurement problem
Pull requests aren’t a perfect proxy for developer productivity, and Anthropic acknowledges this directly. But they argue PRs correlate closely with what engineering teams actually care about: shipping features, squashing bugs, and getting updates to users faster.
The company positions these metrics as complementary to existing engineering KPIs rather than a replacement. Teams can layer them alongside DORA metrics or sprint velocity to spot directional changes after adopting Claude Code.
Available now for paid tiers
Contribution metrics are rolling out to Claude Team and Enterprise customers. Setup requires installing the Claude GitHub App, toggling on GitHub Analytics in admin settings, and authenticating to your organization’s GitHub account. Metrics populate automatically as teams use the tool.
This release follows Anthropic’s January 2026 launch of Claude Code 2.1 and Claude Cowork. The agentic coding tool, which operates directly in developers’ terminals, can handle everything from building features to resolving merge conflicts autonomously. It maintains project awareness and pulls from external sources like Google Drive and Figma via MCP integration.
For teams evaluating AI coding tools, having concrete productivity data could help justify the investment—or reveal whether the hype matches reality.
Image source: Shutterstock
Credit: Source link
