ThoughtSpot MCP Server

What ThoughtSpot MCP Server MCP server does

ThoughtSpot MCP Server MCP server exposes ThoughtSpot data to compatible MCP clients through a set of query and retrieval tools. The server connects to a selected ThoughtSpot instance, manages authentication with OAuth, and returns answers based on live data from that instance.

This is intended for agent workflows where the model needs to ask questions of ThoughtSpot-backed data. The repository describes the server as providing relevant data and live answers, but does not enumerate the individual tool names or query operations in the supplied material.

How it works

An MCP client connects to the ThoughtSpot service and starts an authorization flow. OAuth client details can be created through the project’s client-registration page for hosts that do not support dynamic client registration or require a statically configured client ID and secret.

During registration, an administrator can optionally provide a ThoughtSpot URL. Associating the OAuth client with an instance avoids requiring users to enter that instance URL during authorization. After authentication, the server can query the connected ThoughtSpot instance and return results to the MCP client.

Setup and configuration

For a hosted setup, register a client at the provided ThoughtSpot client-registration page and copy the generated OAuth client ID and secret into the MCP client’s ThoughtSpot connector configuration. The README notes that generated client details are shown only when they are created, so they should be saved at that time.

For local development, install the Node.js dependencies with npm install, copy .dev.vars, and fill it with the ThoughtSpot instance URL and access token. Start the development server with npm run dev. The supplied instructions do not specify the exact environment-variable names used in .dev.vars.

Tools and capabilities

ThoughtSpot MCP Server MCP server supports these documented capabilities:

• Querying data in a ThoughtSpot instance.
• Retrieving relevant data for an agent request.
• Returning live answers from the connected ThoughtSpot data.
• Authenticating access through OAuth.

The codebase organizes tool schemas and implementations separately. Tool definitions are stored in src/servers/tool-definitions.ts, while handlers are implemented in src/servers/mcp-server.ts. A version registry controls which tools are exposed for each API version, including stable and beta entries.

Limitations and notes

The supplied README does not list individual tool names, supported query syntax, response formats, or access-control behavior. It also does not identify exact production endpoint configuration or environment-variable names.

Manual OAuth client registration is relevant only to MCP hosts that lack dynamic client registration or require static OAuth credentials. The README refers to ChatGPT and Claude as examples of MCP clients, but compatibility can depend on each client’s connector and OAuth support.

Changes to the tool set require updates to the version registry, tests, and documentation. New stable tools should be added to the default version, while beta tools belong in the beta entry. ThoughtSpot MCP Server MCP server is therefore best evaluated against the specific tools and API version exposed by the configured registry.