Which agent, model or vendor should do this work?
Compare agents, models and AI vendors on the same kind of work, on the same scale as your engineers. Decide where agents earn their place and what stays human.
| Contributor | Merged | Cost per change | Rework | Cycle time |
|---|---|---|---|---|
| Team baseline (human) | 78% | $41 | 9% | 2.8d |
| claude-refactor | 81% | $14 | 9% | 1.9d |
| codex-refactor | 72% | $11 | 14% | 2.2d |
| cursor-agent | 69% | $17 | 12% | 2.5d |
Every vendor claims their agent is best. Your data should decide.
Vendor benchmarks are not your codebase
Public leaderboards test toy tasks, not your repos, standards and review culture.
Apples and oranges
Agents, assistants and people are measured with different tools, so nobody can compare them.
Renewals on gut feel
Without a head to head comparison, budgets follow the loudest champion.
See which agent wrote what, commit by commit.
AI share by team
AI Checkpoints reads the AI agent recorded in each commit and splits every change into AI and human lines, so you see how much of each team’s work comes from agents.
Checkpoints at a glance
Total checkpoints, average AI attribution and tokens per checkpoint, against the previous period.
Breakdown by team
Users, checkpoints and AI attribution for every team in one table.
AI and human lines
AI changed lines next to human lines added, modified and removed.
| Team | Users | Checkpoints | Avg AI attribution | AI lines | Human lines |
|---|---|---|---|---|---|
| Data | 118 | 612 | 52.8% | 102.4K | +64.1K/15.2K/−9.3K |
| Frontend | 104 | 655 | 51.9% | 109.7K | +71.8K/16.4K/−9.8K |
| Backend | 104 | 521 | 50.6% | 84.9K | +57.2K/12.9K/−7.6K |
| Platform | 86 | 584 | 49.7% | 91.3K | +64.9K/14.6K/−8.7K |
| QA | 60 | 598 | 47.2% | 88.6K | +71.3K/14.8K/−9.1K |
Every agent, side by side
Each checkpoint shows the AI agent behind the commit, the tokens it used and how much of the change it wrote, so Cursor, Claude Code and Codex are compared on your real work.
Agent per commit
See which AI agent produced each change, by repository.
Tokens and attribution
AI tokens and the share of each commit written by AI.
Filter by agent
Narrow down to one agent, team, repo or issue group.
| Checkpoint ID | Repository | AI agent | AI tokens | AI attribution |
|---|---|---|---|---|
| ckpt-7Hq2LmX9vRa4 | Platform | Cursor | 92.4K | 21.6% |
| ckpt-Tn5wB8cKq3Zy | Platform | Claude Code | 118.7K | 18.2% |
| ckpt-Rf4yP1sDv7Xm | Backend | Cursor | 129.3K | 23.4% |
| ckpt-Lk9eG2hNw6Ub | Data | Codex | 84.1K | 26.8% |
One scale for people, assistants and agents.
Live in your stack, not another spreadsheet.
Connect your stack
Link Git (GitHub, GitLab, Bitbucket, Azure DevOps), Jira, CI/CD and your AI coding tools with read-only access.
Waydev maps the work
Every PR, review, ticket and AI session is tied to a team, an initiative and the tool or agent that produced it.
Act on Signals
Waydev Agent flags what changed, suggests Goals, and answers follow up questions in plain language.
Waydev reads engineering metadata through read-only APIs. Source code content is never ingested. Visit the Trust Center
Questions engineering leaders ask
Is this used to rank individual engineers?
How is work grouped by type?
Do you benchmark against other companies?
Trusted by FORTUNE 500 enterprise companies
READY TO ADOPT A DEVELOPER PRODUCTIVITY INSIGHT PLATFORM?
