CBCloudByte PMS
Activity Tracking

Claude Code Usage Analytics & Activity Tracking for Engineering Teams

Capture every Claude Code session, prompt, observation, commit, and machine heartbeat across your team — and turn it into AI, Project, Developer, and Process insights. We caught 24% ghost seats on our own pilot.

The problems

One Capture Layer. Four Things You Can't See Today.

  • You can't see how the AI is actually being used. Prompts, observations, skill invocations, and CLAUDE.md compliance live on each developer's laptop. You don't know which prompts work, which skills go unused, or whose CLAUDE.md is drifting from the team's. (→ AI Insights)
  • You can't tell if AI is moving projects faster. Ticket throughput, DORA metrics, and per-project AI leverage aren't connected. You're spending on Claude Code without a measurable line back to delivery velocity. (→ Project Insights)
  • You can't see who's stuck — until standup tomorrow. 24% of our seats were never activated. Of the rest, a 34× gap between power users (67 sessions/month) and occasional users (2/month). Real-time signal on who's blocked, who's flying, and who needs help. (→ Developer Insights)
  • You can't enforce the process you wrote down. Skills mandated but skipped. Code review rules ignored. CLAUDE.md rules silently drifted. Process compliance is invisible — until something breaks in production. (→ Process Insights)
The capture layer

Four Signals. One 90-Second Install.

One sync agent per machine reads Claude Code's local database, your local git repos, and a process heartbeat — then ships to your org subdomain every three minutes. No VPN, no proxy, no workflow change.

Prompts

Every prompt the developer typed into Claude Code, with project path, git branch, session ID, and timestamp. Searchable in full-text — the raw signal of how engineers are actually working with AI.

Observations

Tool invocations, file reads, file edits, summaries, and skill usage — the full Claude Code session graph. Turns “what did Claude do?” from anecdote into queryable data.

Local Git Commits

Every commit on every local repo (not just Claude sessions) — author, branch, files, full diff, line stats. AI-summarized on ingest by Claude Sonnet 4.6. Ties AI usage back to shipped code.

Machine Heartbeat

Every 3 minutes: agent health, claude-mem DB status, last sync, background job state. Tells you who's actually online, who's drifted offline, and who never installed it.

acme.cloudbyte.ai/dashboard
LIVE
Sessions
668
+12%
Prompts
4,312
+18%
Commits
2,059
+228×
Spend
$842
−6%
Token spend · 7d
By model
MoTuWeThFrSaSu
Recent activity
RMRaj pushed 3 commits2m
PSPriya ran 8 sessions9m
ACAlex synced skills22m
MPMaya updated CLAUDE.md1h
Our 30-day pilot — all four signals, real team

What Activity Tracking Caught On Our Own Team

22 developers, 12 projects, four signals, one month.

0
Developers tracked
12 projects
0
Sessions captured
Mean 30/developer
0
Prompts logged
~6.5 per session
0%
Never-activated
$2,800/mo recovered
For CTOs / VP Eng

Decisions Backed by Data, Not Anecdote

  • Per-seat utilization report for every renewal — fact-based, not anecdotal (Developer Insights).
  • DORA-aligned multipliers per project to prove AI leverage (Project Insights).
  • Process compliance reporting — surface skill and CLAUDE.md drift before it hurts production (Process Insights).
  • 24% unused seats — that catch alone paid for the platform in month one.
For Security / Compliance

Every prompt, accounted for

  • Prompt text, tool output, file paths — exportable per-user for audits.
  • Data isolated per org; four-role RBAC enforced server-side.
  • Self-host on your AWS for full data sovereignty.

Tracks activity across the tools your team uses

Claude Code
GitHub
GitLab
Bitbucket
Linear
Google Workspace
FAQ

Questions engineering leaders ask

Stop Guessing. Start Measuring AI Across Your Team.

A 15-minute demo with a founder. See your team's four-insight profile in your own dashboard.