See every AI tool and agent in use, and what each one can reach.
Classie inventories the AI running across browsers, endpoints and Kubernetes, then maps the users, MCP servers, data and credentials behind each tool. The result is your AI posture, with the gaps flagged.
AI tools arrive faster than security reviews can track them.
of enterprises have employee GenAI tools in production.
a16z CISO Survey, 2026run MCP servers or agent-to-app integrations in production.
a16z CISO Survey, 2026already run autonomous agents with tool access or write permissions.
a16z CISO Survey, 2026of employees use AI tools their employer has not sanctioned.
BlackFog and Sapio Research, 2026Every AI tool, agent and identity, sanctioned or not.
Every GenAI app, agent, MCP server, skill and plugin the sensors observe, with the identity behind it and how it signs in. Sanctioned status comes from your approved list.
| Tool | Status | Account | Users |
|---|---|---|---|
| ChatGPT | Sanctioned | Corporate SSO | 142 |
| ChatGPT | Personal account | Personal | 38 |
| Claude | Sanctioned | Corporate SSO | 61 |
| Perplexity | Unsanctioned | Personal | 24 |
| Otter.ai | Unsanctioned | Personal | 9 |
47 tools and agents · 18 skills and plugins · 12 sanctioned · 6 personal accounts
What each AI tool can reach, and where the gaps are.
Classie links users to AI tools, and AI tools to the data, MCP servers and credentials they reach. Gaps in connectivity, configuration and credentials become findings. Pick a tool to see its paths.
Claude CodeAgent · corporate accountClaude Code is sanctioned, but one unsanctioned MCP server gives it read and write access to the production analytics database.
15 minute installation. No blind spots.
See every tool, agent and connection in your own environment within the first 15 minutes.
Runs in your environment. Your data never leaves the perimeter you govern.