Weave is an engineering intelligence platform. It combines large language models and domain-specific machine learning to understand engineering work: how much work was done, how good it is, and how AI contributed.
Of all the metrics in Weave, code output is the most important. It’s our
algorithm for measuring engineering productivity, unlike lines of code or PR counts, it understands
the meaning and impact of every change.
What Weave measures
- Engineering output and delivery — code output, PR cycle time, review turnaround, task delivery, and deploy metrics, by team, person, or repository
- Code quality — review depth, code turnover, reverts, and bugs introduced
- AI coding tools — adoption, AI-written code, cost, and output for tools like Claude Code, Cursor, Codex, and GitHub Copilot, plus coding agents (AI tools overview)
- AI return on investment — AI spend next to the output it produced, in the AI ROI report
- Where engineering time goes — work by project and initiative, including R&D capitalization
Connect your data
Weave reads from the tools your team already uses:
Start with connecting your data, then explore the reports.
- The Weave MCP server (Model Context Protocol) lets Claude, Cursor, and other MCP clients query your Weave metrics directly.
- Weave Router is one managed endpoint that picks the right model for every request from your coding agent.
Last modified on October 17, 2026
