Agents

Agents lists every AI agent observed across your estate, and the skills, MCP servers, system prompts, models and people behind each one. Open it from Observe > Agents.

Source chips at the top show which sources contribute agents, for example Azure AI Foundry, Copilot Studio, Endpoint discovered, Endpoint runtime, MCP Gateway and OpenAI Compliance. The headline tiles are Agents observed, Observed calls, People and MCP reach (with blocked calls).

Agents

One row per application or agent: Source, Capabilities (observed and declared), Calls, Users, Governance and Last seen. Use search and Filters to narrow the list. Click a parent application to filter the list to its child agents.

Skills

Which skills agents use, and how widely. This tab is read-only usage evidence. To install, configure or remove skills, use the Skills Library.

MCP servers

MCP servers agents call: Used by, Evidence, Calls, Users, Tools and Blocked. Use it to see which agents reach which servers, and where calls are being blocked.

System prompts

The system prompts observed for agents, with their size in tokens, potential compression savings, risk, and the top suggestion for each. Use it to find oversized or risky prompts. To act on compression savings, see Costs & Savings.

Users

People who run agents: the agents each person used, calls, blocked calls and devices.

Agent drawer

Click an agent to open its drawer. It has five tabs: Overview, Users, Dependencies & Calls, Activity and System prompt. The Overview shows the agent's identity, type, parent applications, sources, top users and source instances.