Every AI dollar, and what it bought you.
Know exactly what you’re spending on AI tooling, how it maps to usage, and whether that investment is translating into real productivity gains.
- 3.2×Return on AI tooling spend
- +31%More PRs merged per engineer
- −38%Cost per merged PR, year on year
Multi-layered attribution
Weave uses a hierarchy of methods, so every AI contribution gets attributed.
Attribution method · falls back automatically
Verified · native API
Direct integration with tools like Cursor and Copilot. The strongest signal available.
Inferred · commit metadata
Git co-authorship on commits, for tools that expose no API.
Estimated · time-series inference
Modeled from activity windows. Directionally accurate, always labeled as estimated.
What usage tracks
Three signals that tell you not just what’s being used, but whether it’s working.
Foreground vs. Background AI
Usage categorized into foreground AI (interactive IDE assistants) and background AI (autonomous agents). Distinguish active pairing from automated assistance.
What spend and ROI covers
Four signals that connect what you pay for AI to what you get back.
Subscription spend · monthly
Sample data| Tool | Seats | Monthly | Per seat |
|---|---|---|---|
| Cursor | 120 | $4,800 | $40 |
| Claude Code | 86 | $3,440 | $40 |
| Windsurf | 44 | $1,320 | $30 |
| Amazon Q | 30 | $600 | $20 |
Weave tracks per-user costs for Amazon Q, Claude Code, Cline, Cursor, Firebender, Sourcegraph, and Windsurf. For tools without per-user tracking, costs are calculated at the org level.
Track subscription costs across every AI coding tool your org uses. See monthly spend per tool, cost per active user, and how spending trends over time. Configure your subscriptions once in Weave settings and cost analysis updates automatically.
Connect AI spend to actual engineering output. See what portion of output is AI-assisted, how AI usage correlates with productivity, and where you’re getting real efficiency gains versus where you’re just adding cost.
See who’s actively using AI tools, which tools are most adopted, and how adoption is changing over time. Break it down by team, tool, and individual.
Two kinds of AI contributor
Weave distinguishes between two fundamentally different categories of AI contributor.
Foreground AI
Interactive tools engineers use directly
Claude Code, Cursor, Windsurf, tools where an engineer is actively pairing with AI to write or review code.
Background AI
Autonomous agents working independently
Devin, Bugbot, CodeRabbit, agents that execute tasks, open PRs, and fix bugs without direct human involvement.
Autonomous task tracking
Multi-file refactoring, bug fixes, task completion.
ROI tracked separately
Cost and output measured against human-driven work.
Agent ROI and cost
Agent activity is tied to the same cost tracking and output metrics as human-driven AI. See the specific return on your autonomous agent stack.
Cost per merged PR
MCP integration
Weave’s MCP server gives agents observability into your team’s health. Connected agents can query velocity metrics, drill into PRs, and align generated code with team standards.
How it works
No manual tracking. No spreadsheets built from memory at the end of the quarter.
Built for engineering and finance
Weave’s R&D Capitalization report identifies which development tasks qualify for capitalization by analyzing PR output, task association, and project data.
Year-to-date
The report defaults to a year-to-date view, giving finance teams a consistent snapshot of development activity across the fiscal period.
Team, repo, or project
Break the report down by whichever unit your finance team files against.
Slice capitalization data precisely so the numbers accurately reflect the specific work being performed. Pairs with Jira or Linear for the most accurate results.
Cross-reference AI spend with output and adoption data in one view.
Know whether your investment is working, and where to adjust.