generated: '2026-08-09'
method: derived
source: json-schema/mcp360-tool-schemas.json + openapi/*.json + probed response envelopes
shape: tool-registry, not a resource CRUD model
note: MCP360 is a gateway, not a system of record. There are no addressable resources, no ids in any response,
  and no $ref graph between tool schemas — every one of the 38 specs has a flat components.schemas of
  per-tool input objects with zero cross-references. The graph below is the operating model derived from
  the tool registry and the response envelope, not an ERD over stored entities.
counts:
  services: 38
  tools: 106
  rest_operations: 178
  cross_schema_refs: 0
entities:
- name: Service
  description: A namespaced bundle of tools mounted on the gateway (e.g. google-search, whois). 38 exist.
  source: GET https://connect.mcp360.ai/api/v1/{service}
  fields:
  - service (slug, the path segment)
  - name
  - description
  - version
  - isPremium
  - creditsPerUse
  identifier: service slug
- name: Tool
  description: One callable capability inside a Service. 106 exist.
  source: GET https://connect.mcp360.ai/api/v1/{service} -> tools[]
  fields:
  - name
  - description
  - endpoint
  - method
  - inputSchema (JSON Schema)
  - outputSchema (JSON Schema)
  identifier: service + name; the REST operationId is "{service}_{name}" (POST) and "{service}_{name}_get"
    (GET twin)
- name: Execution
  description: One tool invocation and its result envelope.
  source: POST /api/v1/{service}/{tool} response
  fields:
  - success (bool)
  - data (object, shaped by the tool outputSchema)
  - metadata.service
  - metadata.tool
  - metadata.executionTime
  - metadata.creditsUsed
  - metadata.timestamp
  identifier: none — no execution id or request id is returned
- name: Credit
  description: The billing unit. Consumed per successful execution; failed calls are not charged.
  source: https://mcp360.ai/pricing + Execution.metadata.creditsUsed
  fields:
  - creditsUsed (per execution)
  - creditsPerUse (per service)
  - monthly allowance (per plan)
  identifier: not exposed via the API
- name: ApiKey
  description: 'The credential presented as X-API-Key or Authorization: Bearer.'
  source: https://dashboard.mcp360.ai (Settings -> API Keys)
  fields:
  - key (opaque)
  - 'state: active | inactive | banned (inferred from the 401 message)'
  identifier: not exposed via the API
- name: Project
  description: Plan-scoped grouping (1 on Free, 2 Starter, 10 Professional, unlimited Advanced).
  source: https://mcp360.ai/pricing
  fields:
  - name
  - members
  identifier: not exposed via the API
relationships:
- from: Service
  to: Tool
  type: has_many
  via: tools[]
  evidence: GET /api/v1/{service} embeds the full tools array.
- from: Tool
  to: Service
  type: belongs_to
  via: endpoint path segment /api/v1/{service}/{tool}
- from: Tool
  to: Execution
  type: has_many
  via: POST {endpoint}
- from: Execution
  to: Tool
  type: belongs_to
  via: metadata.tool
- from: Execution
  to: Service
  type: belongs_to
  via: metadata.service
- from: Execution
  to: Credit
  type: has_one
  via: metadata.creditsUsed
- from: Service
  to: Credit
  type: has_one
  via: creditsPerUse
  note: Flat rate per service; 1 credit on every one of the 38 services captured.
- from: ApiKey
  to: Execution
  type: has_many
  via: X-API-Key request header
  note: Not modelled in any response; inferred from the auth contract.
gaps:
- No execution id or request id is returned, so a call cannot be referenced after the fact.
- Credits, keys and projects are dashboard-only — there is no account or usage API.
- Tool output schemas are published at /api/v1/{service} but are NOT carried into the generated OpenAPI,
  whose 200 response types data as a bare object.