[ 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 workflow changeNo new toolsNo timesheetsPatented capture
IDEVS Code · JetBrains
TerminalShell sessions
AgentsClaude Code · Codex
One recordrefactor-billing
Prompts12
Sessions6 · 41 min
Drafts discarded2
Pull requestMerged
RECONCILED
AI invoicethis month
Claude Code$420
Copilot$55
Sessions matched6/6
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.

01

Collects

Every session and delivery, captured at the source.

02

Analyzes

Cost, effort, and quality connected per delivery item.

03

Evaluates

Leverage vs waste, team by team.

AITRAX

Every session, captured at the source.

IDE, terminal and agent sessions, recorded where the work happens and normalized into one record per delivery. Only encrypted activity telemetry leaves the machine.

WHAT AITRAX RECORDED · TODAY
Sessions, by source
SOURCEWORKTIMETOKENSSTATUS
VS CodeIDEinvoice.ts · tax.ts14 minCAPTURED
Claude CodeAGENTrefactor-billing41 min25.6k tokCAPTURED
TerminalSHELLnpm test -- rounding2 minCAPTURED
JetBrainsIDETaxService.java27 minCAPTURED
CodexAGENTadd-webhook-retry19 min8.2k tokCAPTURING

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

Security & deployment
[ 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.

HOW ONE DELIVERY WAS SCORED, HOUR BY HOUR
Claude CodeDev testingCopilot
3.1×value per $
Worth it. Value delivered exceeds cost.
$475total AI spend
$420Claude Code
$55Copilot
$120flagged as waste

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

Get your AI report in 7 days
[ 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.

TEAM A$2,00050 points
TEAM B$10,00050 points
AI spendexcess spendvalue points
SAME OUTPUT, 5× THE COST

Team A returned 5× the value per dollar.

TEAM A$2,00050 points
TEAM B$2,000500 points
AI spendvalue points
SAME SPEND, 10× THE OUTPUT

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.

Executive presenting results on a large chart to a leadership team in a glass boardroom
01 · EXECUTIVES

Which investment to make first.

Engineering lead leaning in to review code on a developer's monitor
02 · ENGINEERING LEADERS

Which teams need support, which tools earn their keep.

Developer with headphones working across two monitors of code
03 · 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.

Security & deployment

Get your AI report
in 7 days.

Installs in hours. Insights from day one. A defensible baseline in a week.

Agent