What AI costs, and what it built.
AI spend by person, team, space and epic, tied to the commits and pull requests that did the work — and the issues they belong to. Your tools send the data in; the app never calls out.
The real module, running right here.
Working software on sample data — the same interface that renders inside Jira. It plays a short tour when it comes into view; click anywhere in it to take over.
Every number here comes from the app's own report code, run on sample activity. Send in a GitHub push or a Claude Code report and it is read by the same code that reads them inside Jira, then every figure is recomputed. Spend from a vendor's bill is shown as spend; Claude Code's own cost counter is marked as an estimate. Installed, your tools send the data in and the app never calls out.
Built for the job, not the brochure.
Spend where it lands
AI spend by person, team, space and epic, with spend from keys nobody is linked to shown as unassigned rather than hidden.
Claude Code reports itself
Claude Code sends its own usage over OpenTelemetry every five minutes: cost, tokens, sessions, lines, commits and pull requests. Metrics only — no prompts, no code.
Commits tied to issues
GitHub pushes and pull requests are tied to issues by the keys in their message, title or branch name. Work signed by Claude Code, Copilot, Cursor, Codex, Gemini or Devin counts as AI-assisted.
One person, many accounts
A person’s Jira account, email, GitHub login and AI accounts are joined into one contributor. Your corrections stay put.
Adoption and cycle time
Who uses AI lightly, regularly or heavily, and the cycle time of AI-assisted issues next to the rest.
Who built this
An issue panel shows the people, commits, pull requests and AI tools behind each issue, its hours and cycle time, and its share of AI cost.
Three steps from install to answer.
Switch on telemetry
Roll a few settings out to Claude Code through your managed settings file. It starts reporting within five minutes.
Add a GitHub webhook
Pushes and pull requests arrive signed with a secret; keys in branch names tie them to issues.
Read the overview
Spend, people using AI, unassigned spend and cycle time, by team, with every figure traceable to what was sent.
Inside Atlassian, with Jira’s rules.
Nothing is fetched. Claude Code, GitHub and your own script send data to the app; your AI vendors’ admin keys never reach it. Only Jira administrators, and the people they name, can see any of it. What “never leaves your instance” means →
- Spend from a vendor’s bill is shown as spend; Claude Code’s own cost counter is marked as an estimate
- Telemetry is metrics only — no prompts, no code, no file contents
- GitHub deliveries are checked against the webhook secret
- Anthropic and OpenAI billing through a script you run with your own keys in testing
Straight answers.
How does it see AI spend without leaving Atlassian?
Your tools send the data in; the app never calls out. Claude Code reports its own usage over OpenTelemetry, GitHub posts commits and pull requests by webhook, and a small script you run in your own GitHub Actions reads your Anthropic and OpenAI billing with admin keys that never reach the app. The billing script is still being tested against live Anthropic and OpenAI accounts.
What does Claude Code send?
Usage metrics: cost, tokens, sessions, lines changed, commits and pull requests, per person. The settings the app gives you send metrics only — no prompts, no code, no file contents.
Who can see it?
Jira administrators, and the people they name. Nobody else sees any of it.
Guides and articles.
See Claude Code spend per person, team and Jira issue
Point Claude Code’s own telemetry and a GitHub webhook at Workforce Intelligence, and see what AI coding costs — per person, per team, and per Jira issue it helped build.
Read the guide →Turn a support address into Jira issues with Cloudflare Email Routing
Forward a support address on your own domain into Jira, so each new message becomes an issue and each reply lands as a comment — without giving any app a mailbox.
Read the guide →Earned value for Jira projects, in plain English
Spent 48% of the budget — is that good? Earned value answers by comparing what you spent with what that money was supposed to buy. Here is the whole idea, with one worked example.
Read the guide →Be first in when it opens.
One email when early access opens. Nothing else — no newsletter, no drip sequence, and the list isn't shared with anyone.
Early access members get the first year at launch pricing, locked.
Your address is stored with Resend, the email service that will send that one email.