specification: API Commons MCP
specificationVersion: '0.1'
provider: CData
providerId: cdata
generated: '2026-09-05'
method: searched
source: >-
  https://docs.cloud.cdata.com/en/API/MCP.md and
  https://docs.cloud.cdata.com/en/API/MCP-Management.md (published tool lists),
  https://docs.cloud.cdata.com/en/API/MCP-API.yaml (the provider's own OpenAPI for the MCP
  transport), plus a live POST probe of https://mcp.cloud.cdata.com/mcp on 2026-09-05.
status: published
name: CData Connect AI MCP
description: >-
  CData Connect AI ships a hosted, remote MCP server that exposes every configured data
  source connection as MCP tools. Two servers are published: a data server at /mcp
  (catalog/schema/table/column discovery plus SQL query and stored-procedure execution) and
  a Management server at /mcp/mgmt (list available sources, create/test connections, manage
  toolkits). Both speak the MCP Streamable HTTP transport.
deployment:
  mode: remote
  endpoint: https://mcp.cloud.cdata.com/mcp
  auth: oauth
  verified: probed
  note: >-
    OAuth 2.1 (authorization code + PKCE S256, dynamic client registration) OR HTTP Basic
    with a Connect AI user id and Personal Access Token, per
    https://docs.cloud.cdata.com/en/API/MCP.md. `auth: oauth` records the primary,
    agent-usable path; the Basic alternative is captured in authentication/.
    No stdio package is published — there is no `npx`/`uvx` install for this server, so an
    agent reaches it over HTTPS or not at all.
endpoints:
- name: Connect AI MCP (data)
  url: https://mcp.cloud.cdata.com/mcp
  transport: streamable-http
  docs: https://docs.cloud.cdata.com/en/API/MCP.md
  openapi: openapi/cdata-mcp-api-openapi.yml
- name: Connect AI Management MCP
  url: https://mcp.cloud.cdata.com/mcp/mgmt
  transport: streamable-http
  docs: https://docs.cloud.cdata.com/en/API/MCP-Management.md
  note: >-
    Included in the free Developer Edition (described there as a Developer Edition
    exclusive alongside the data server).
- name: Connection-scoped MCP (Embed)
  url: https://mcp.cloud.cdata.com/mcp/connections/{connectionId}
  transport: streamable-http
  docs: https://docs.cloud.cdata.com/en/API/mcp-connections.md
  openapi: openapi/cdata-mcp-api-embedded-openapi.yml
  note: Connect AI Embed surface; routes MCP traffic to a single data source connection.
probe:
  url: https://mcp.cloud.cdata.com/mcp
  method: POST
  body: '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
  http_status: 401
  www_authenticate: >-
    Bearer resource_metadata="https://mcp.cloud.cdata.com/.well-known/oauth-protected-resource/mcp",
    scope="openid profile email"
  body_returned: '{"error":"UNAUTHORIZED","message":"Missing or invalid Authorization header"}'
  checked: '2026-09-05'
  interpretation: >-
    The server is live and RFC 9728-correct: it refuses anonymous tools/list and hands back a
    resource_metadata pointer that itself resolves (200, saved to
    well-known/cdata-mcp-oauth-protected-resource.json). The live tool inputSchemas are
    therefore auth-gated. The tool list below is NOT introspected — it is transcribed from
    CData's own published documentation and from the `ops` enum in the provider's MCP-API.yaml.
    Full JSON Schema for each tool requires authenticated introspection.
oauth:
  authorization_server: https://mcp.cloud.cdata.com
  protected_resource_metadata: https://mcp.cloud.cdata.com/.well-known/oauth-protected-resource
  authorization_endpoint: https://mcp.cloud.cdata.com/authorize
  token_endpoint: https://mcp.cloud.cdata.com/token
  registration_endpoint: https://mcp.cloud.cdata.com/register
  pkce: S256
  scopes: [openid, profile, email, offline_access]
  mcp_protocol_version: '2025-06-18'
tools:
  source: >-
    Names and descriptions from https://docs.cloud.cdata.com/en/API/MCP.md; the operation
    base names come from the `ops` query-parameter enum in the provider's own MCP-API.yaml.
    Input schemas not captured — the live endpoint is auth-gated (see probe above).
  data_server:
  - name: getCatalogs
    op: get_catalogs
    kind: read
    description: Retrieve a list of available connections (catalogs) from Connect AI.
  - name: getSchemas
    op: get_schemas
    kind: read
    description: Retrieve database schemas for a specific catalog.
  - name: getTables
    op: get_tables
    kind: read
    description: Retrieve tables and views for a specific catalog and schema.
  - name: getColumns
    op: get_columns
    kind: read
    description: Retrieve column definitions for a specific catalog, schema and table.
  - name: getProcedures
    op: get_procedures
    kind: read
    description: Retrieve stored procedures for a specific catalog and schema.
  - name: getProcedureParameters
    op: get_procedure_parameters
    kind: read
    description: Retrieve parameter metadata (names, types, directions) for a stored procedure.
  - name: getInstructions
    op: get_instructions
    kind: read
    description: >-
      Source-level semantic guidance the base skill requires an agent to call before any
      schema/table/column call. Present in the MCP-API.yaml ops enum; the base skill
      (skills/cdata-connect-ai-base.md) makes it the first step of the discovery workflow.
  - name: queryData
    op: execute_sql
    kind: write
    description: >-
      Execute SQL against the connected data sources. The ops enum also exposes narrower
      execute_select / execute_insert / execute_update variants.
  - name: executeProcedure
    op: execute_procedure
    kind: write
    description: Execute a stored procedure for the chosen catalog and schema.
  tool_filtering:
    note: >-
      The MCP endpoint accepts `tools` (universal|sql|source) and `ops` query parameters that
      narrow what tools/list returns. Narrowing-only — they cannot enable tools the server
      has disabled. An empty or unrecognised value returns HTTP 400.
      Source: openapi/cdata-mcp-api-openapi.yml.
  management_server:
  - name: create_connection
    kind: write
    description: >-
      Save a new data source connection. Requires name, source and a properties object keyed
      by the field names returned by get_source_properties. OAuth sources return a sign-in URL.
  - name: get_source_properties
    kind: read
    description: >-
      Describe the configuration fields a data source accepts, split into basic and advanced,
      with allowed values and conditional follow-up fields.
  - name: list_available_sources
    kind: read
    description: List the data sources the user can connect to, with internal source name and category.
  - name: list_connections
    kind: read
    description: List saved connections with id, name and source type.
  - name: test_connection
    kind: write
    description: Verify a saved connection by attempting to connect to the data source.
  note: >-
    The Management MCP documentation also describes Toolkit Management tools
    (add_tool_to_toolkit and related). Only the five Connection Management tools above were
    read in full from the published page and are recorded here; the toolkit tools are named
    in the docs but their parameters were not transcribed.
clients:
  documented: [Claude, Microsoft Copilot Studio, n8n, Databricks, Gemini Enterprise, ChatGPT Directory]
  source: https://docs.cloud.cdata.com/en/Clients/AITools.md
maintainers:
- FN: Kin Lane
  email: info@apievangelist.com