← Back to Discover

CloudByte PMS

A dashboard that records how developers use Claude Code, Cursor and GitHub Copilot, then reports usage, token spend, adoption, security events and searchable project knowledge for the whole team.

Observability & evalsAI security & guardrails
Visit website ↗

What it does

CloudByte AI works through a small sync agent installed on every developer's computer. The agent looks at the session records that Claude Code, Cursor and GitHub Copilot keep locally, leaves them untouched, gathers git commits from the repositories you choose, and pushes the lot to your organisation's dashboard every few minutes. The site says developers change nothing about how they work and that dashboards are live within about ten minutes. The dashboard breaks activity down into sessions, prompts, tool calls and tokens, per developer, project and model. It shows spend with a cache-hit ratio, an estimated return on investment, and paid seats that have no activity. Each commit also gets an AI-written summary next to its file list and diff. On the security side, the site says prompt scanning catches credentials, secrets and personal data before they leave the developer machine, and that unapproved plugins and MCP servers are detected, with an audit trail. Prompts, commits, tickets and session summaries are indexed into a searchable knowledge base for each project. Only session metadata and commit metadata leave the machine; the site says no source code files are uploaded. Sign-in uses Google SSO with domain restriction or magic links, with four-tier role-based access. The Enterprise plan lets you bring your own Anthropic key, and the site says it can be self-hosted or run air-gapped.

Capabilities

  • Tracks sessions, prompts, tokens and commits from Claude Code, Cursor and GitHub Copilot
  • Spend split by developer, project and model, with inactive paid seats flagged
  • AI summaries of each commit and a searchable knowledge base per project
  • Prompt scanning for secrets plus detection of unapproved plugins and MCP servers
  • Google SSO and four-tier role-based access
  • Enterprise plan adds your own Anthropic key and self-hosted or air-gapped setups

When to use it

  • You lead an engineering team and want to see who actually uses the paid AI coding seats
  • You need token spend attributed to developers, projects and models
  • You want a check for leaked secrets in prompts and for AI tools nobody approved
  • You want project context to carry over between developers instead of living on one laptop