Edit

Inspect a local agent with the Agent Inspector

Important

Items marked (preview) in this article are currently in public preview. This preview is provided without a service-level agreement, and we don't recommend it for production workloads. Certain features might not be supported or might have constrained capabilities. For more information, see Supplemental Terms of Use for Microsoft Azure Previews.

The Agent Inspector is a browser-based UI for poking at a Microsoft Foundry agent running on your local machine. It connects to the local agent's HTTP / SSE endpoint, shows requests and responses, and lets you replay messages while you iterate on prompts, tools, and code.

Prerequisites

The inspector ships in its own extension (azure.ai.inspector) and is installed automatically by the microsoft.foundry meta-package. It's also pulled in as a dependency of azure.ai.agents, so installing just the agent extension is enough.

Note

The inspector only targets a local agent on http://localhost:8088 by default. It doesn't authenticate to or proxy a deployed Foundry agent. For the deployed agent, use azd ai agent invoke and azd ai agent monitor.

Start the agent locally

  1. Start the agent in one terminal:

    azd ai agent run

    By default, azd ai agent run opens the inspector for you when the local server starts listening.

  2. To skip the auto-launch, pass --no-inspector:

    azd ai agent run --no-inspector

Launch the inspector manually

If you started the agent yourself, for example with python app.py or dotnet run, or if you closed the inspector tab and want to reopen it, launch the inspector manually.

  1. Run the launch command:

    azd ai inspector launch

    The UI is served on http://localhost:8087 by default and connects to an agent on http://localhost:8088.

  2. Configure both ports if needed:

    # Agent on port 9000, UI on port 9001.
azd ai inspector launch --port 9000 --inspector-port 9001

  3. To seed the inspector with an existing conversation or session ID for replaying a specific run, pass the IDs explicitly:

    azd ai inspector launch \
  --conversation-id conv_abc123 \
  --session-id sess_xyz789

Inspect local traffic

When the inspector is open, the single-page app streams the agent's responses over SSE and mirrors them in your terminal as well. You can:

  • Inspect the full request and response payloads for each turn.
  • Continue an existing conversation by reusing the conversation ID across turns.
  • Reset to a fresh session at any time.

The inspector doesn't modify your agent code or agent.yaml. It's purely a runtime view.

Related content

  • Last updated on