Arthur MCPArthur MCP
Dashboard
How it worksFeaturesDocsEarly accessGitHubDashboard

How to Configure Response Limits

Cap how large a tool's response can get before it reaches the AI, so one oversized upstream response can't blow out the AI client's context.

Steps

  1. Open the server and go to its Settings tab.

    Open the capture server whose response-limit settings will be demonstrated

  2. Find Response Limits and enable it.

    Open the Settings tab, where the Response Limits panel is located

    Ensure Response Limits is enabled so the four configuration fields are available, including when recovering from an interrupted run

  3. Set the limits you need:

    • Max response length — the maximum size (in characters) of the JSON sent to the AI, from 1,000 to 10,000,000.

    • Max depth — how many levels of nested JSON are kept, from 1 to 50.

    • Array slice — the maximum number of items kept in any array, from 1 to 10,000.

    • Error truncate length — the maximum size of an error message, from 100 to 1,000,000.

      Set the maximum number of response characters sent to the AI

      Set how many levels of nested JSON are preserved

      Set the maximum number of items retained in each array

      Set the maximum number of characters retained from an error body

      Restore the fixture's baseline maximum response length

      Restore the fixture's baseline maximum JSON depth

      Restore the fixture's baseline maximum array size

      Restore the fixture's baseline error-body length

      Disable Response Limits through its checkbox to return the capture fixture to its baseline state

  4. Changes save automatically a moment after you stop typing.

Confirm it worked

Call a tool that normally returns a large or deeply nested response — the response the AI receives is now truncated/limited according to your settings.

Good to know

  • This limits the JSON payload itself (length, nesting, array size), not a raw HTTP byte cap — it's applied when Arthur maps the upstream response before handing it to the AI.
  • Useful for APIs that return large lists or deeply nested objects that would otherwise flood the AI's context window.

Related

Tutorial video