generated: '2026-09-06' method: searched status: published source: https://docs.dremio.com/dremio-cloud/ai-integration/mcp-server docs: - https://docs.dremio.com/dremio-cloud/ai-integration/ - https://docs.dremio.com/dremio-cloud/ai-integration/mcp-server - https://docs.dremio.com/current/developer/mcp-server deployment: mode: both endpoint: https://mcp.dremio.cloud/mcp/{project_id} install: git clone https://github.com/dremio/dremio-mcp.git && uv sync && uv run dremio-mcp-server run package: https://github.com/dremio/dremio-mcp auth: oauth verified: probed note: >- Two distinct products. The Dremio-hosted remote server is an https endpoint an MCP client POSTs to today — mcp.dremio.cloud (US control plane) or mcp.eu.dremio.cloud (EU), with the project id as the final path segment. The self-hosted server is an open-source Python project cloned from GitHub and run under uv with a personal access token; it is not on PyPI or npm as a published package, so there is no `npx`/`pipx` one-liner. server: name: dremio transport: http url: https://mcp.dremio.cloud/mcp/{project_id} regions: - control_plane: US endpoint: https://mcp.dremio.cloud/mcp/{project_id} - control_plane: EU endpoint: https://mcp.eu.dremio.cloud/mcp/{project_id} self_hosted: repository: https://github.com/dremio/dremio-mcp runtime: python >= 3.11 via uv auth: personal access token modes: - id: FOR_DATA_PATTERNS description: Explore data, generate SQL from natural language, and create views in Dremio. - id: FOR_SELF description: System introspection and performance analysis. - id: FOR_PROMETHEUS description: Enrich insights with Dremio metrics from a connected Prometheus setup. authentication: hosted: scheme: oauth2 authorization_server: https://login.dremio.cloud protected_resource: https://mcp.dremio.cloud/mcp scopes: - dremio.all - offline_access dynamic_client_registration: https://login.dremio.cloud/oauth/register pkce: S256 discovery: - https://mcp.dremio.cloud/.well-known/oauth-protected-resource - https://mcp.dremio.cloud/.well-known/oauth-authorization-server note: >- Claude.ai and ChatGPT connect with the endpoint alone (dynamic client registration); other clients need a Native OAuth application created in the Dremio console. self_hosted: scheme: personal-access-token config_file: ~/.config/dremioai/config.yaml authorization_model: >- Tools execute as the authenticated Dremio user — the docs state that Dremio's existing access controls are enforced and agents operate with the same permissions as that user. live_introspection: attempted: '2026-09-06' endpoint: https://mcp.dremio.cloud/mcp method: 'POST {"jsonrpc":"2.0","id":1,"method":"tools/list"}' http_status: 401 www_authenticate: Bearer resource_metadata="https://mcp.dremio.cloud/.well-known/oauth-protected-resource/mcp" result: gated note: >- The live tool list and its inputSchema require an authenticated OAuth session. The RFC 9728 challenge is well-formed and the referenced resource metadata resolves (HTTP 200), so the server is real and reachable; only the schema is gated. tools: - name: GetUsefulSystemTableNames description: List the Dremio system tables worth querying for introspection. source: https://docs.dremio.com/dremio-cloud/ai-integration/mcp-server input_schema: unknown-auth-gated - name: GetSchemaOfTable description: Return the schema of a table in the Dremio catalog. source: https://docs.dremio.com/dremio-cloud/ai-integration/mcp-server input_schema: unknown-auth-gated - name: RunSqlQuery description: Execute a SQL query against Dremio and return results. source: https://docs.dremio.com/dremio-cloud/ai-integration/mcp-server input_schema: unknown-auth-gated tools_note: >- Only the three tools the documentation names verbatim ("You should see dremio listed with available tools like GetUsefulSystemTableNames, GetSchemaOfTable, and RunSqlQuery") are recorded. The docs say "like", so this is a partial list, and the full set plus every inputSchema needs an authenticated tools/list. Nothing here is guessed.