# How it works — Captured where the work happens. Scored by what it delivers.

CodeTogether captures every AI-assisted session at the point of work, connects it to the delivery it produced, and scores what that delivery was worth to your business. Nothing changes for developers.

## Where the data comes from

The data only exists where the work happens. Git and ticketing tools see the result. CodeTogether captures the process: every prompt, session, and draft, live in the IDE, the terminal, and agent sessions, then reconciled against your actual AI invoice. You can't get this from GitHub after the fact.

No timesheets. No workflow change. Patented capture.

- Sources: IDE (VS Code, JetBrains), Terminal (shell sessions), Agents (Claude Code, Codex)
- One record per delivery: prompts, sessions, drafts discarded, pull request merged
- Reconciled against the AI invoice: Claude Code $420, Copilot $55, sessions matched 6/6, total $475

## AITrax · the data collection engine

CodeTogether sits alongside. No change to the work. CodeTogether's data collection engine, AITrax, runs beside the IDEs and coding agents on the workstation, collecting and normalizing data on how each delivery is created, automatically. Only encrypted activity telemetry leaves the machine, into your own environment.

1. **Collects** — every session and delivery, captured at the source.
2. **Analyzes** — cost, effort, and quality connected per delivery item.
3. **Evaluates** — leverage vs waste, team by team.

Only encrypted activity telemetry leaves the machine, into your SaaS tenant or your own cluster.

## Value Points · the measurement model

Scored from the work itself. Every delivery is scored against your business priorities, not an industry benchmark. So you can finally answer three questions.

- **Speed — Is it better and faster?** Delivered at 73 hours, review approved, no rework.
- **Return — Is the cost worth it?** $475 of AI spend, 3.1× value per dollar. Worth it.
- **Direction — Is it aligned to your goals?** Scored against the priorities your business set.

One delivery, scored hour by hour: 3.1× value per $. $475 total AI spend ($420 Claude Code, $55 Copilot). $120 flagged as waste.

AI slop scores zero, and every score is challengeable from the pull request itself.

## What the analysis finds

The analysis uncovers what was invisible.

- **Waste found — AI work dropped on the floor.** Sessions and drafts that never reached a pull request. Counted, not hidden.
- **Churn detected — Spinning without progress.** AI retrying hard on an inefficient approach. Visible, not buried.
- **Multiplier found — Your best patterns, identified.** The workflows and people producing outsized return.

## Finding · Team A vs Team B

Once you can measure value, the differences jump out.

- **Same output, 5× the cost.** Team A: $2,000 AI spend, 50 value points. Team B: $10,000, 50 value points. Team A returned 5× the value per dollar.
- **Same spend, 10× the output.** Team A: $2,000, 50 value points. Team B: $2,000, 500 value points. Team B delivered 10× the value output.

This is the ROI lens for your own organization: it tells you what to amplify, and what to stop.

## What each role gets

From score to next action.

- **Executives** — which investment to make first.
- **Engineering leaders** — which teams need support, which tools earn their keep.
- **Developers** — which work is worth improving next.

Every finding arrives attached to the person who can act on it, in the language that role already uses.

## Get your AI report in 7 days

Installs in hours. Insights from day one. A defensible baseline in a week. Book the call: https://codetogether.com/get-started

---

- Human view of this page: https://codetogether.com/how-it-works
- All pages as Markdown: https://codetogether.com/llms.txt
- Contact: info@codetogether.com

CodeTogether · Patent pending