generated: '2026-09-04'
method: derived
status: candidate
source: openapi/bot-butcher-classification-api-openapi.yml
description: >-
  CANDIDATE MCP tool surface for the Bot Butcher Classification API, derived from the two published
  OpenAPI operations. Bot Butcher does NOT ship an MCP server: no hosted endpoint, no stdio package,
  no mention of MCP anywhere on its site, and no MCP package in any registry. This manifest is a
  proposal for what a server would expose if one existed, and is pointed at from apis.yml with
  type X-MCPServerCandidate — never MCPServer — so it is not read as a shipped agent surface.
deployment:
  mode: none
  endpoint: null
  install: null
  package: null
  auth: api-key
  verified: probed
search_evidence:
  - url: https://api.botbutcher.com/mcp
    method: POST tools/list
    status: 404
    finding: No MCP endpoint on the API host.
  - url: https://botbutcher.com/mcp
    status: 404
    finding: No MCP endpoint on the primary site.
  - url: https://botbutcher.com/.well-known/ai-plugin.json
    status: 404
    finding: No plugin manifest.
  - url: https://registry.npmjs.org/-/v1/search?text=botbutcher
    status: 200
    finding: No package, therefore no stdio server to install.
  - url: https://botbutcher.com/documentation
    status: 200
    finding: The reference documents two REST operations and no agent or MCP integration.
auth_if_built:
  type: api-key
  header: x-api-key
  note: >-
    A server would need to hold a per-form Bot Butcher API key; keys are static and long-lived with
    no documented rotation, so an operator should scope one key per form and per deployment.
tools:
  - name: classify_message
    description: >-
      Classify a contact form message as spam or legitimate, optionally within the context of the
      website it was posted on.
    rest:
      operationId: classifyMessage
      method: POST
      path: /
    inputSchema:
      type: object
      required:
        - message
      properties:
        message:
          type: string
          description: The text to classify as spam or not.
        domain:
          type: string
          description: Root domain where the contact form is located; boosts accuracy.
        metadata:
          type: object
          additionalProperties: true
          description: Arbitrary data to store with the request.
    consequence: write
    note: >-
      Billable and non-idempotent — every call creates a new classification and, unless storage is
      disabled on the account, a new stored message. There is no reversal operation.
  - name: get_message
    description: Retrieve a previously classified message and its verdict by message_id.
    rest:
      operationId: getMessage
      method: GET
      path: /message/{message_id}
    inputSchema:
      type: object
      required:
        - message_id
      properties:
        message_id:
          type: string
          description: Unique identifier of a previously submitted message.
    consequence: read