Lightweight AI work intelligence for small and AI-native engineering teams.
Turn local AI coding sessions into a unified team dashboard — see how your team actually uses Cursor, Claude Code, OpenClaw, and other agents, without uploading code or transcripts.
- Why DevPulse
- Quick Start
- Daily Workflow
- Automatic Sync
- Dashboard Views
- Supported AI Tools
- CLI Commands
- Privacy
- Roadmap
- License
Engineers rarely stick to one AI coding tool. A typical day might span Cursor for quick edits, Claude Code for multi-file refactors, and other agents for specialized tasks. Each tool ships its own usage dashboard — but none of them show the full picture across tools, projects, or teammates.
Standups, Slack threads, and Jira tickets capture intent — not how work was actually done with AI. For small engineering teams and AI-native teams especially, there is no lightweight way to:
- See whether everyone is on the same page about daily progress
- Understand how AI fits into real delivery (not just token counts)
- Spot uneven adoption or runaway costs before they become a problem
- Align when the team is AI-native — most work runs through local agents, but visibility still lives inside each tool's silo
DevPulse fixes this with a simple loop:
Local AI tools → devpulse CLI → Team dashboard
(Cursor, Claude Code, OpenClaw) (usage, summaries, trends)
- Open the dashboard and sign up
- Create a team or join one with an invite code
- Share the invite code with teammates
Requires Node.js 22+
npm install -g devpulse-aiInstall from this repo instead
cd cli
npm install && npm run build && npm linkdevpulse loginOpens your browser to authorize the CLI — similar to Claude Code device login. Each laptop only needs to do this once.
devpulse syncThe CLI scans local AI tool logs, extracts lightweight metadata, and uploads only what's needed for the team dashboard. Unchanged sessions are skipped automatically.
Check pending work without uploading:
devpulse status- Go to Dashboard → Overview
- Explore team activity, per-member breakdowns, token trends, and daily AI summaries
| When | What to run |
|---|---|
| First time on a machine | devpulse login |
| After a coding session | devpulse sync |
| Check pending work | devpulse status |
| Set-and-forget | devpulse schedule install --every 6 or --daily --at 09:00 |
Most teams either run
devpulse syncat end of day, or install a background schedule for automatic uploads.
Schedule devpulse sync to run in the background:
# Every 6 hours
devpulse schedule install --every 6
# Once per day at 9:00 AM (local time)
devpulse schedule install --daily --at 09:00
# Check or remove the schedule
devpulse schedule status
devpulse schedule remove| Platform | Scheduler |
|---|---|
| macOS | launchd |
| Linux | cron |
Logs are written to ~/.devpulse/schedule.log.
| View | What it shows |
|---|---|
| Overview | Team activity, token trends, tool/model breakdown, daily summary |
| Members | Per-person sessions, active time, and what they worked on |
| Sessions | History of synced sessions with project, agent, model, and summary |
| Settings | CLI tokens — revoke devices or create manual tokens |
The CLI auto-detects whichever tools are installed on your machine. Missing tools are skipped.
| Tool | Token usage in dashboard |
|---|---|
| Claude Code | Full |
| Codex | Full |
| OpenClaw | Full |
| Cursor | Sessions & activity (tokens not available locally) |
| Command | Description |
|---|---|
devpulse login |
Authorize this machine via browser. --no-browser to paste a token manually. |
devpulse sync |
Upload new/changed sessions. --dry-run to preview, --force to re-sync all. |
devpulse status |
Show login state, detected tools, and pending sessions. |
devpulse schedule install |
Auto-sync on a timer. --every <hours> or --daily [--at HH:MM]. |
devpulse schedule status |
Show the installed schedule. |
devpulse schedule remove |
Uninstall auto-sync. |
More CLI details → cli/README.md
DevPulse uploads derived metadata only — never full transcripts, prompts, or source code.
What is uploaded per session:
- Tool, model, timestamps, message count
- Token usage (where the local tool exposes it)
- Project name and a hashed project path (raw paths are never sent)
- A short local summary for daily recap generation
What is NOT uploaded:
- Full conversation transcripts
- Prompts or source code
- Raw filesystem paths
Config and sync state live in ~/.devpulse/ on your machine.
- GitHub commit history — correlate AI sessions with commits and PRs
- Team lead review & scoring — review daily summaries and leave lightweight feedback
- Effective usage patterns — surface which workflows and tools correlate with the best outcomes
MIT