Thoughtspot

Install

Works with Claude Code, VS Code, Cursor, Windsurf. Read the repository before you give any tool access to your data.

npx agentmarket install connector/thoughtspot-app-mcp-server

About

Use cases: data-analysis, data-visualization

No reviews yet. Installed it? Reviews come from verified installers only. How reviews work

Permissions

What this server can do and what it asks for, from its tool list and install config. Checked Oct 25, 2026.

Tools

14 read · 6 write · 0 delete

Secrets

1 (API keys, tokens or passwords)

Install scripts

Nothing to install (hosted)

Runs as

Hosted (remote endpoint)

Tools (20)

Read from the source code; the running server may differ. Descriptions are the server’s own words; hidden characters are shown as [U+…].

• check_connectivityCheck Connectivityread-only

  Ping tool to test connectivity and authentication. You can use this if other tool calls are failing to verify if the connection is working.

• create_analysis_sessionCreate Analysis Sessionread-only

  Start an analytical session with the Analytics Agent. This is the first step in a three-step workflow: create a session, send a message, then poll for updates. Once created, you can use the returned `analytical_session_id` to send analytical questions via `send_session_message` and retrieve answers via `get_session_updates`. Sessions are conversational, so you can ask follow-up questions in the same session without creating a new one. Using a single analytical session is preferable, because it reuses the same data source selection. When the user's intent is to create a dashboard, use the first…

• create_dashboardCreate Dashboard

  Create a dashboard from a list of answers, allowing the user to revisit the results later. You can use this if the user asks for a dashboard or liveboard, or asks to save the results from the analysis. This can be a useful way to save the results to revisit later, or present them to other users.

• create_model_sessionStart Model Session

  Start a session with the ThoughtSpot model builder, to build a NEW data model or to EDIT an existing one. Returns a model_session_id used by send_model_message, get_model_updates and finalize_model. - NEW model: pass `connection_identifier`, the warehouse connection to build on. Ask the user which connection to use and confirm the exact GUID with them before calling this; it should not be guessed, reused from a previous model, or assumed from a default. - EDIT an existing model: pass `model_identifier` instead (confirm which model with the user first). Creating a session does not build anythin…

• createLiveboardCreate Liveboard from Answers

  Create a liveboard from a list of answers

• finalize_modelSave Model

  Save the model. Call first WITHOUT confirm to get a summary for review, then let the user decide: to make more changes, go back to send_model_message; to save, call again with confirm=true to persist and get the URL. confirm=true is the only step that saves. Don't finalize until the model is actually built — if the builder has only returned a plan (an mrd update) and nothing has been built, there is nothing to save.

• get_dashboard_statusGet Dashboard Statusread-only

  Continue waiting for a dashboard that is still being built or changed. Call this when `create_dashboard` or `modify_dashboard` returned `status: "in_progress"`, passing the `task_id` they gave you. Each call waits a short while before returning, so calling it repeatedly is the correct way to wait. Keep calling until `status` is `completed`, `needs_input` or `failed`. Do not start a new `create_dashboard` or `modify_dashboard` call for the same work, or you will create duplicate dashboards. Check `status`: `completed` (give the user `dashboard_url`), `in_progress` (call `get_dashboard_status`…

• get_dataGet Dataread-only

  Fetch the full data (columns and rows) of a saved Answer or Liveboard, identified by its GUID — typically an object found via `search_objects`. Use this to explain, describe, summarize, analyze, or interpret what a ThoughtSpot object, Liveboard, Answer, or visualization actually shows. When the user wants to explain, describe, summarize, analyze, or interpret what a specific object contains or shows, an accurate answer comes from this tool's data, so you should fetch it first with the object's `id` and answer from the returned rows. Runs the object's saved question (its existing filters and co…

• get_model_updatesBuilding Model…read-only

  Get the model builder's new updates for the instruction in progress. Call it repeatedly until is_done is true. A full build takes 1–2 minutes, so an empty batch with is_done=false means it is still working, not stalled. React to the update types: - `mrd` — a proposed plan (Model Requirements Document). This is a HARD STOP: show the plan to the user and WAIT for their explicit approval (or edits). NEVER auto-approve it or send a 'build it'/'yes' message yourself. (If a build starts without any `mrd` — a backend config that skips the plan step — tell the user the plan step was skipped rather tha…

• get_session_updatesGet Analysis Session Updatesread-only

  Get the latest updates from the Analytics Agent. You can call this after `send_session_message` to retrieve the Agent's response. If `is_done` is false, you can call this tool again to continue polling, as the Agent is still generating a response. Even if `is_done` is false, you can use the updates to show status updates or progress to the user, so that they are informed about the ongoing process. An empty `session_updates` list while `is_done` is false is normal; it means the Agent is still thinking. When `is_done` is true, the Agent has finished and the results in `session_updates` are compl…

• getAnswerGet Answer for a Questionread-only

  Get the answer to a question from ThoughtSpot database

• getDataSourceSuggestionsGet Data Source Suggestions for a Queryread-only

  Get data source suggestions for a query. Use this tool only if there is not datasource id provided in the context or the users query. If mulitple data sources are returned, and the confidence difference between the top two data sources is less than 0.3, ask the user to select the most relevant data source. Otherwise use the data source with the highest confidence to get the relevant questions and answers for the query.

• getRelevantQuestionsGet Relevant Questions for a Queryread-only

  Get relevant data questions from ThoughtSpot database

• list_orgsList Orgsread-only

  List the Orgs the authenticated user can access on the ThoughtSpot instance, including the ID, name, and description of each Org. The Org marked `is_active: true` is the one currently active for your account, which all tool calls operate against. Use this to tell the user which Org they are in. Only available when authenticated via OAuth. If the list is large , summarize or show only the most relevant orgs (e.g. the active one and a few others) rather than listing all of them, unless the user explicitly asks to see all orgs.

• modify_dashboardModify Dashboard

  Change an existing dashboard. It can rearrange and restyle what is already there, add new charts, add short explanatory text, and organise charts into groups or tabs. `dashboard_id` is always required. Add `task_id` as well to continue something you already started, which is how you answer a question the designer asked you or send a follow-up instruction about the same dashboard. Pass the change in `instructions` in the user's words. If it is too vague to act on, the designer comes back with a question. New charts use the data sources the dashboard already has unless you say otherwise. Chan…

Signals

Stars

34

Forks

12

Open issues

2

Contributors

11

Last commit

Oct 20, 2026 (9 days ago)

Last release

Sep 17, 2025 (1 year ago)

From GitHub (fetched Oct 24, 2026).

Details

Registry name

app.thoughtspot/mcp-server

Version

1.0.1

Auth

Auth not stated, Bearer token

License

License unknown

Repository

github.com/thoughtspot/mcp-server

First seen

Oct 24, 2026

Last checked

Oct 24, 2026

Source

Official MCP Registrygithub-graphql