Skip to main content
Weave connects to the AI coding tools your engineers use, including Claude Code, Cursor, GitHub Copilot, Codex, Windsurf, and Devin, and measures adoption, AI-attributed code output, and spend across people, teams, and tools. It labels tools as foreground AI, which engineers interact with directly, or background AI, which works automatically in the background, so work from autonomous agents and review bots can be told apart from work engineers do with an assistant. For per-developer telemetry from Claude Code, Cursor, GitHub Copilot, Codex, or Grok Build, follow the Weave plugin installation guide.

Supported tools

Direct integrations

These tools can be connected directly to Weave:

Cline

OpenTelemetry integration for code generation tracking

Cursor

API integration for IDE usage and code generation

Kiro

S3 integration for AI IDE usage and code generation

Windsurf

API integration for IDE usage and code generation

Claude Code

Telemetry and API integration for code generation and cost

Oh My Pi

OTLP integration for model, token, session, and OpenRouter cost tracking
Other supported tools: Router attribution uses GitAI’s measured code contributions and matching Router session metadata. A native session ID is stable while the client stays on the router; turning the router off starts a new session, so a match attributes the whole GitAI session. Unmatched or conflicting contributions keep their Codex or Claude Code label; using /fm through the Router does not create a separate AI tool. Code attribution does not change token usage or billing records.

Auto-detected tools

These tools are automatically detected from commit metadata (no setup required):
  • CodeRabbit
  • Continue
  • Cubic
  • Devin (connect the Cognition API for ACU activity and cost)
  • Gemini Code Assist (connect the BigQuery integration for IDE usage metrics)
  • Graphite
  • Greptile
  • Jules
  • Junie
  • Nori
  • OpenCode
  • OpenHands
  • Parameter
  • Pi (requires git attribution)
  • Sazabi
  • Warp
  • Wispbit

Foreground vs background AI

Weave categorizes AI tools into two groups: Foreground AI — Tools that engineers interact with directly:
  • AI Gateway, Claude Code, Cline, Codex, Copilot, Cursor IDE
  • Firebender, Gemini CLI, Gemini Code Assist, Grok, Hermes Agent, Junie, Nori, OpenClaw, OpenCode, OpenRouter
  • Pi, Warp, Windsurf, Augment, Auggie Interactive
  • Sourcegraph, Amazon Q, Kiro, JetBrains AI Assistant
  • Weave Router
Background AI — Tools that work automatically in the background:
  • Auggie Automation, Augment Code Review
  • CodeRabbit, Continue, Cubic
  • Cursor Agent, Cursor Bugbot
  • Devin, GitHub, Graphite
  • Greptile, Jules, Junie Bot
  • OpenHands, Parameter, Sazabi, Wispbit
This distinction helps you understand both active AI usage and passive AI assistance.

What you can measure

Once connected, you can track:
  • AI output percentage — What portion of output is AI-attributed
  • AI LOC — Lines of code written with AI assistance
  • AI tool costs — Spending on AI tools, using provider-reported billing where available (including OpenRouter-reported cost from Oh My Pi), or estimated / seat-based otherwise
  • AI efficiency index — A 0 to 100 score that combines AI output, AI usage percentage, AI cost per user, and code turnover
  • AI adoption trends — How AI usage changes over time

Getting started

1

Choose your tools

Select which AI tools your team uses from the supported list
2

Connect integrations

Follow the setup guide for each tool (Cline, Cursor, Kiro, Windsurf, Claude Code, or Oh My Pi)
3

Wait for sync

Allow time for initial data sync (usually a few minutes)
4

View metrics

Access AI metrics in Reports and Overview pages

Data privacy

Weave tracks AI usage metadata and prompt content:
  • Number of AI-generated lines
  • Tool usage frequency
  • Cost and subscription data
  • Commit metadata with AI attribution
  • User prompt text (used for activity classification; system messages are excluded)
We never access:
  • Source code content or diffs
  • AI model responses
  • File paths or project structure
  • Individual keystroke data

Common questions

Which AI coding tools does Weave support?

Claude Code, Cursor, GitHub Copilot, Codex, Windsurf, Devin, Cline, Kiro, Amazon Q, Augment, Gemini CLI, Gemini Code Assist, Sourcegraph, OpenRouter, and the other tools listed under supported tools. Tools such as CodeRabbit, Greptile, Jules, and OpenHands are detected from commit metadata with no setup.

How does Weave separate agent work from developer-assisted work?

Tools are grouped into foreground and background AI. Add the Background vs. foreground AI chart to a custom dashboard to split AI lines of code between the two, and register internal bots as custom AI agents with the category that fits.

Do I need the Weave plugin?

For many tools, no: usage and cost come through an API key, telemetry, or your code platform. The Weave plugin for Claude Code, Cursor, GitHub Copilot, Codex, or Grok Build adds per-developer session data that AI skills scoring uses.

Can I compare what each tool costs with the output it produces?

Yes. AI tool costs shows spend per tool and, where the data allows, per engineer. The AI report’s Cost per output by tool chart divides that spend by AI-attributed code output; see AI ROI.
Last modified on October 17, 2026