Skip to content

MCP Inspector

MCP Inspector is a browser-based developer tool for testing and debugging MCP servers — similar to Postman but for the Model Context Protocol. Use it to call tools interactively, browse resources, preview prompts, and inspect raw protocol messages without writing any client code.

Launch the Inspector

No installation required. Run it directly with npx:

npx @modelcontextprotocol/inspector

The UI opens at http://localhost:6274. A proxy server also starts on port 6277 to handle protocol bridging.

Connect to misp-workbench

Get a token

First generate a scoped MCP token from the /mcp/config endpoint:

GET /mcp/config
Authorization: Bearer <your-api-token>

Copy the Bearer token value from the response — you'll paste it into the Inspector.

Configure the connection

In the Inspector's left sidebar:

  1. Set Transport to Streamable HTTP
  2. Set URL to http://localhost:8000/mcp (or your instance URL)
  3. In the Authorization field, enter Bearer <mcp-token>
  4. Click Connect

The Inspector completes the MCP handshake (initialize → notifications/initialized) and shows the server capabilities.

Local dev without auth

If MCP_AUTH_ENABLED=false (the default for local dev), skip the token — leave the Authorization field empty and connect directly.

What you can do

Tools tab

Lists all tools with their input schemas. Select a tool, fill in the parameters using the auto-generated form, and click Call to execute it. The raw JSON response is shown in the output pane.

Useful for verifying tool behavior before connecting an AI client:

  • Test search_events with Lucene queries to confirm index connectivity
  • Call detect_indicator_type on a batch of values to validate type detection
  • Run enrich_indicator to confirm a module is enabled and returning results
  • Trigger create_event_report to verify write access

Resources tab

Lists all misp:// resources. Select one and click Read to fetch its content. Useful for checking that the galaxy and taxonomy submodules are correctly initialised.

Prompts tab

Lists all analysis prompt templates. Select a prompt, fill in the arguments (e.g. keyword: ransomware), and click Get Prompt to preview the message that would be sent to an LLM.