generated: '2026-08-30'
method: searched
source: >-
  https://docs.textql.com/core/how-it-works/mcp/ana-as-mcp-server,
  https://docs.textql.com/core/how-it-works/mcp/client-setup,
  https://docs.textql.com/core/how-it-works/mcp/tool-groups
name: Ana MCP Server
status: published
deployment:
  mode: remote
  endpoint: https://app.textql.com/mcp
  auth: oauth
  verified: probed
x-evidence:
  fetched: '2026-08-30'
  url: https://app.textql.com/mcp
  method: POST
  body: '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'
  http_status: 401
  response_body: 'no bearer token'
  www_authenticate: >-
    Bearer resource_metadata="https://app.textql.com/.well-known/oauth-protected-resource/mcp"
  note: >-
    The live tool schemas are auth-gated. The 401 is RFC 9728-correct: it names the protected-resource
    metadata document for this exact resource, which resolves (200) and declares the authorization
    server and the mcp:tools scope. Tool names, descriptions and grouping below are taken verbatim
    from TextQL's published MCP documentation; their inputSchemas require an authenticated
    tools/list call and are NOT recorded here rather than being guessed.
transport: Streamable HTTP
oauth:
  authorization_server: https://app.textql.com
  protected_resource_metadata: https://app.textql.com/.well-known/oauth-protected-resource/mcp
  resource: https://app.textql.com/mcp
  scopes_supported:
  - mcp:tools
  grant_types: [authorization_code, refresh_token]
  pkce: S256
  dynamic_client_registration: https://app.textql.com/oauth/register
  spec: RFC 9728 protected-resource metadata + RFC 8414 authorization-server metadata
clients_documented:
- Claude (web)
- Claude Code
- Cursor
- Windsurf
- ChatGPT
- Microsoft Copilot
- MCP Inspector
configuration:
  toolset_scoping:
    query_param: toolsets
    header: X-MCP-Toolsets
    keywords: [all, default]
    example: https://app.textql.com/mcp?toolsets=analysis,catalog,sandcastle
  dynamic_mode:
    query_param: dynamic
    example: https://app.textql.com/mcp?dynamic=1
    meta_tools: [list_available_toolsets, get_toolset_tools, enable_toolset, disable_toolset]
    note: >-
      In dynamic mode the server tracks the session's active sandcastle, so sandcastle tools may be
      called without passing a sandbox_id. Requires a client that supports MCP tool-list updates.
tool_groups:
- name: analysis
  default: true
  description: Ask the data analyst agent questions, synchronous and streaming.
  tools: [ana, ana_ask, ana_poll]
- name: catalog
  default: true
  description: Discover the connectors you can query.
  tools: [list_connectors]
- name: sandcastle
  default: false
  description: >-
    Run code and queries in a managed Python sandcastle — create, inspect and stop sandcastles; run
    Python in a persistent kernel; load connector data or saved .tql queries from the Ontology into
    dataframes; run shell commands; read and upload files; and propose Ontology changes as a
    reviewable patch.
  tools: [execute_python, run_query]
  tools_note: >-
    Only execute_python and run_query are named explicitly in the published docs; the rest of this
    group is described in prose and its exact tool names are not published.
- name: ontology
  default: false
  description: >-
    Review and govern proposed Ontology changes — list pending patches, read one, and approve, deny
    or restore.
tools:
- name: ana
  group: analysis
  description: >-
    Ask any natural-language question. Ana connects to the warehouse through the caller's configured
    connectors and Ontology, runs the analysis, and returns the answer plus a link to the full chat
    thread in TextQL. Synchronous; can take up to a minute.
- name: ana_ask
  group: analysis
  description: >-
    Streaming variant for clients with short request timeouts. Starts the analysis and returns
    immediately.
- name: ana_poll
  group: analysis
  description: >-
    Streams progress, answer text, charts and report links until the run started by ana_ask completes.
- name: list_connectors
  group: catalog
  description: List the data connectors the caller can target, so a question can be scoped to a source.
- name: execute_python
  group: sandcastle
  description: Execute Python in a persistent kernel inside a managed sandcastle.
- name: run_query
  group: sandcastle
  description: Run a query in the sandcastle, loading results as a dataframe.
returns:
  - The answer, grounded in the caller's warehouse data
  - A thread link back into the TextQL UI for the full chat, SQL and audit trail
  - Resource links for charts, dashboards and reports
access_management:
  docs: https://docs.textql.com/core/how-it-works/mcp/managing-access
  note: Users and organization admins can review and revoke authorized MCP clients.
outbound_mcp:
  note: >-
    TextQL is also an MCP CLIENT — the Public RPC API's MCPService lets an organization register
    external MCP servers (Notion, GitHub, internal APIs) into Ana's context, with an OAuth flow per
    server. See openapi/textql-platform-api-openapi.yml, paths under
    /textql.rpc.public.mcp.MCPService/.
  operations:
  - MCPService_GetMCPServers
  - MCPService_UpsertMCPServers
  - MCPService_ToggleMCPServer
  - MCPService_DeleteMCPServer
  - MCPService_InitiateOAuthFlow
  - MCPService_HandleOAuthCallback
  - MCPService_ClearOAuthToken