ThoughtSpot

The ThoughtSpot MCP Server provides secure OAuth-based authentication and a set of tools to query ThoughtSpot data from your instance. It can run remotely and supports multiple transports, making it easy to connect your favorite LLMs or AI tools to ThoughtSpot data.

How to use

You connect your MCP client to ThoughtSpot by configuring an MCP server URL and choosing an authentication method. After you establish the connection, ThoughtSpot datasources appear under your resources. Select a datasource to set the context for your queries, then ask analytical questions and let the MCP tools fetch relevant data, answers, or liveboards from ThoughtSpot.

How to install

Prerequisites: Node.js version 18 or higher and npm. Ensure you have access to a ThoughtSpot instance with the necessary permissions.

# 1) Clone the MCP server repository (or navigate to your project workspace)
git clone https://example.com/repo/mcp-server.git
cd mcp-server

# 2) Install dependencies
npm install

# 3) Start the development server (local development workflow)
npm run dev

Additional sections

Configuration and usage details are shown below. You can run the MCP server locally (stdio transport) or connect to the remote MCP endpoint. The following are explicit, practical options you can use right away.

MCP server connection methods shown in examples:

• Remote MCP URL (HTTP transport) to connect directly to the server.

Configuration and usage notes

HTTP transport (remote) configuration is used to connect to the MCP server via an HTTP endpoint. The following URL is used to access the MCP API remotely.

{
  "type": "http",
  "name": "thoughtspot",
  "url": "https://agent.thoughtspot.app/mcp",
  "args": []
}

Self-hosted and stdio transport (local)

If you run the MCP server locally, you can use the stdio transport to connect through your environment. Use the following configuration to start the local MCP server with npm and environment variables.

{
  "type": "stdio",
  "name": "thoughtspot_stdio",
  "command": "npx",
  "args": ["@thoughtspot/mcp-server"],
  "env": {
    "TS_INSTANCE": "<your ThoughtSpot Instance URL>",
    "TS_AUTH_TOKEN": "<ThoughtSpot Access Token>")
  }
}

Stdio transport example specifics

When using stdio transport, specify the final runtime command exactly as shown to launch the server locally. The example uses npx to run the MCP server package and passes your ThoughtSpot instance URL and access token as environment variables.

{
  "type": "stdio",
  "name": "thoughtspot_stdio",
  "command": "npx",
  "args": ["@thoughtspot/mcp-server"],
  "env": {
    "TS_INSTANCE": "https://my-thoughtspot-instance.thoughtspot.cloud",
    "TS_AUTH_TOKEN": "YOUR_TS_AUTH_TOKEN"
  }
}

Security and access

Use OAuth for authenticating as yourself to access your ThoughtSpot data. If you need to connect from tooling that requires a bearer token, you can pass the token in the appropriate headers as shown in API examples.

Troubleshooting and notes

If you encounter OAuth errors due to CORS or SAML settings, add agent.thoughtspot.app to your ThoughtSpot security configurations under CORS whitelisted domains and SAML redirect domains. For server errors or authorisation failures, collect client logs and review timestamps to identify the issue.

Endpoints and tools overview

The MCP server exposes a set of tools to query and manipulate ThoughtSpot data. You can ping the server to test connectivity, fetch relevant questions, retrieve answers, construct liveboards, and get datasource suggestions. These tools work with the configured MCP transports.

Available tools

ping

Test connectivity and authentication with the MCP server.

getRelevantQuestions

Retrieve relevant questions based on a user query from ThoughtSpot analytics.

getAnswer

Fetch the answer to a specific question from ThoughtSpot analytics.

createLiveboard

Create a liveboard from a list of answers.

getDataSourceSuggestions

Suggest ThoughtSpot data sources relevant to a query.