turbopuffer-mcp
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@turbopuffer-mcpquery the 'documents' namespace for similar items"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
turbopuffer-mcp
A Turbopuffer MCP server built with dedalus-mcp and DAuth-style credential exchange.
This server follows the same DAuth north-star pattern as x-api-mcp:
Credentials are provided by clients at runtime.
The server declares a
ConnectionwithSecretKeys.Tool calls dispatch through Dedalus secure connection handles.
Features
DAuth-compatible API key authentication (
TURBOPUFFER_API_KEY).Configurable Turbopuffer region/base URL via
TURBOPUFFER_BASE_URL.Read + write tool coverage for the core Turbopuffer API.
Includes a smoke ping tool for quick MCP handshake validation.
Setup
Create a Turbopuffer API key.
Copy env template:
cp .env.example .envRequired:
TURBOPUFFER_API_KEY(provided by the MCP client via DAuth credentials in production flows)
Optional:
TURBOPUFFER_BASE_URL(defaults tohttps://gcp-us-central1.turbopuffer.com)DEDALUS_AS_URL(defaults tohttps://as.dedaluslabs.ai)HOST(defaults to127.0.0.1)PORT(defaults to8080)
Run
uv run python src/main.pyTool Surface
turbopuffer_list_namespacesturbopuffer_get_namespace_metadataturbopuffer_get_namespace_schematurbopuffer_update_namespace_schematurbopuffer_writeturbopuffer_queryturbopuffer_multi_queryturbopuffer_explain_queryturbopuffer_delete_namespaceturbopuffer_cache_warmturbopuffer_measure_recallturbopuffer_export_documents(deprecated Turbopuffer endpoint compatibility)smoke_ping
Notes
Turbopuffer has both
/v1and/v2endpoints. This server mirrors the live docs/API split:Namespace listing/metadata/schema/cache-warm/recall/export use
/v1.Write/query/delete/explain use
/v2.
The export endpoint is deprecated in Turbopuffer docs in favor of paging query APIs, but is still exposed here for compatibility.
Related MCP Connectors
Operate your Vector Panda collections: search and upsert vectors, schema, measured indexes, exports.
A collaborative substrate over your data: vector, knowledge graph, SQL, geospatial, streaming.
- mcpOAuthcom.vibgrate
Query your team's drift, vulnerability, and upgrade data from any AI assistant. OAuth 2.1, 51 tools.
Ingest, manage, and retrieve documents for RAG-powered AI applications