Tinybird MCP Server – Real-Time Analytics Tools
Tinybird provides a fully managed remote MCP server for each workspace, enabling AI agents to query data, inspect data sources, and call published API Endpoints through the Model Context Protocol. Use it when an agent needs a conversational interface over Tinybird analytics without building custom tool wrappers.
Overview
Tinybird is a real-time analytics platform built around data sources, SQL pipes, and published API Endpoints. Its remote MCP server lets AI agents connect directly to a Tinybird Workspace so they can answer questions over analytics data, execute permitted queries, and use published Endpoints as callable tools.
The MCP server is hosted by Tinybird at https://mcp.tinybird.co and is designed for agents that need live workspace context rather than static documentation. It is useful for data exploration, analytics assistants, and prototypes where an LLM should work with Tinybird data through the standard Model Context Protocol.
What the MCP server enables
The Tinybird MCP server exposes Endpoint tools and core tools. Each published API Endpoint in a Workspace can become an MCP tool named after the Endpoint. Those Endpoint tools accept the same parameters as the Endpoint, return the same response as direct API calls, default to CSV output, support JSON, and respect Endpoint rate limits and Token scopes.
Core tools include list_endpoints, list_datasources, list_service_datasources, execute_query, explore_data, and text_to_sql. These tools allow an agent to discover available analytics resources, generate SQL from natural-language questions, run SQL through the Tinybird Query API, and ask exploratory questions over data.
When to use it
Use Tinybird MCP when an AI agent needs access to analytics data, a conversational interface over Data Sources or Endpoints, or a fast way to prototype data-analysis workflows. It is also useful when existing Tinybird API Endpoints should become agent tools without writing custom wrappers.
Direct Tinybird APIs remain a better fit for production applications with fixed query patterns, lowest-latency requirements, or custom caching, retries, observability, and request handling.
Connection and authentication
Tinybird officially supports Streamable HTTP for its remote MCP server. The documented MCP URL is https://mcp.tinybird.co?token=TINYBIRD_TOKEN, where the token is a Tinybird Token. Clients that do not support Streamable HTTP can run mcp-remote as a bridge.
Authentication and available tools depend on Token scopes. Resource-scoped static Tokens or JWTs should be used to limit which Endpoints or Data Sources an agent can access. Admin Tokens expose all available tools and should be reserved for trusted internal tools only.
Key considerations
Tinybird Tokens may include the Tinybird API host. Older Tokens can require a host query parameter. JWTs do not contain region information, so non-default regions require an explicit regional API host in the MCP URL. MCP requests include from=mcp, allowing teams to monitor agent queries using Tinybird service Data Sources.
Supported Transports
streamable_http
URL: https://mcp.tinybird.co?token=${TINYBIRD_TOKEN}
stdio
Command: npx
Args:
-ymcp-remotehttps://mcp.tinybird.co?token=${TINYBIRD_TOKEN}
Frequently Asked Questions
- When should an AI agent use the Tinybird MCP server?
- Use it when an agent needs to explore Tinybird analytics data, call published API Endpoints, answer questions over Workspace data, or prototype a conversational analytics workflow without writing custom Tinybird API wrappers.
- What does the Tinybird MCP server add to an AI agent's capabilities?
- It gives the agent live access to permitted Tinybird Workspace resources, including published Endpoint tools and core tools for listing endpoints, listing data sources, generating SQL, executing queries, and exploring data.
- What can the AI agent access or manage through Tinybird MCP?
- Access depends on the supplied Token. With appropriate scopes, the agent can use published API Endpoints as tools, list Data Sources and service Data Sources, run SQL through the Query API, use text-to-SQL, and ask exploratory questions over data.
- How is authentication configured for the Tinybird MCP server?
- Authentication is configured by passing a Tinybird Token in the MCP URL as `token`. Tinybird recommends resource-scoped static Tokens or JWTs to restrict access, while admin Tokens should only be used for trusted internal tools.
- Which transport should be used for the Tinybird MCP server?
- Use the hosted Streamable HTTP endpoint when the MCP client supports it. For MCP clients that do not support Streamable HTTP directly, Tinybird documents using the local `mcp-remote` stdio bridge with `npx`.