AI Code Pulse Tracker

A lightweight background daemon that tracks your Claude Code AI usage and sends it to your team's analytics backend — so you always know how many tokens your engineering work actually consumes and what it costs.

AI Code Pulse is an AI usage analytics dashboard inside Jira that tracks Claude Code usage alongside rework rate and PR cycle time. It analyzes token consumption, cost, model usage, and cache efficiency, and correlates these signals with engineering activity such as code rework across commits and pull requests, as well as repositories, files, authors, epics, and Jira tasks, to provide visibility into development efficiency and the impact of AI-assisted workflows.

The platform helps engineering teams understand:

• AI usage and costs
• Token consumption
• Model usage
• AI-assisted development patterns
• Code rework and churn
• Pull request lifecycle metrics

AI Code Pulse correlates Claude Code activity with commits, repositories, pull requests, Jira issues, epics, and engineering workflows to provide visibility into how AI tools are used across the organization.

Quick start

Global install

npm install -g aicp-tracker
aicp-tracker setup    # enter your email and Claude plan
aicp-tracker start    # launch the background daemon

Local install (per project)

npm install aicp-tracker
npx aicp-tracker setup
npx aicp-tracker start

That's it. The daemon runs silently and picks up every new session automatically.

Commands

| Command | Description | |---|---| | aicp-tracker setup | Interactive first-time configuration | | aicp-tracker start | Start the background daemon | | aicp-tracker stop | Stop the daemon | | aicp-tracker status | Show daemon status, project path, and pending records | | aicp-tracker flush | Manually send any queued records |

What gets tracked

For each Claude Code conversation turn, the daemon records:

• Token counts — input, output, cache creation, cache read (including ephemeral variants)
• Model — which Claude model was used
• Git branch — branch name at the time of the session, with Jira task keys extracted automatically
• File paths — which files were written or edited in each turn
• Web usage — web search and web fetch request counts
• Timestamps and session IDs — for timeline and per-session analysis

All data is written to a local write-ahead log first, so nothing is lost if the network is unavailable. Records are retried on the next cycle.

Configuration

Run aicp-tracker setup to configure:

• Email — your work email (the one you use for GitHub or Bitbucket), used to attribute usage to your account
• Plan — your Claude subscription tier, used to calculate token costs

{
  "email": "[email protected]",
  "plan": "Pro (monthly)"
}

Plans: Pro (monthly), Pro (yearly), Max, Team (standard), Team (premium), Enterprise

Project path detection

The daemon automatically locates the Claude Code session folder that matches your project by walking up from the configured projectPath until it finds a matching entry under ~/.claude/projects/. This works even when Claude Code names the project by a parent workspace folder rather than the exact directory.

Privacy

Only token counts, model names, file paths, git branch names, and timestamps are collected. Prompt content and code are never read or transmitted.

Requirements

• Node.js 18+
• Claude Code installed and in use