verify_claim

Inputs

{
  "type": "object",
  "properties": {
    "claim_id": {
      "type": "integer"
    },
    "evidence": {
      "type": "array",
      "items": {
        "type": "object"
      },
      "description": "Each: {source_url, quote, alignment_signal}. alignment_signal enum is locked."
    },
    "mode": {
      "type": "string",
      "enum": [
        "agent",
        "byo_key"
      ]
    }
  },
  "additionalProperties": false,
  "required": [
    "claim_id",
    "evidence"
  ]
}

Outputs and behavior

Record your agent's evidence and alignment signals for an existing claim. Use after extract_claims; use get_claim to inspect evidence without changing it. Replaces the stored evidence and updates review status. Signals describe support, contradiction, mixed evidence or inconclusiveness, never a truth verdict; the user judges the claim. The agent performs research, and BYO model execution is not implemented.

The registry does not publish an output schema for this tool. The description above is its documented behavior; response fields beyond it are not specified here.

Example call

Example tool name and arguments, generated from the input schema. Replace placeholder strings and sample numbers with your own values. This is a request template, not a recorded call or response.

{
  "name": "verify_claim",
  "arguments": {
    "claim_id": 0,
    "evidence": [
      {}
    ]
  }
}

Clients

This tool is available to MCP clients using stdio or http.

The registry records transport support, not individual client test results. See the MCP setup reference for named clients, configuration, and tool-list limits.