How is this different from connecting ThoughtSpot MCP directly?

Connected directly, an agent sees every tool the server exposes and holds the credential itself. Through TrustGate you choose which tools each agent gets, and every call is logged against the agent that made it.

What if the ThoughtSpot MCP API changes?

We track the ThoughtSpot MCP schema, so the tool catalogue stays current. Tools you have already allowed keep working, and anything new stays denied until you enable it.

Can I give different agents different ThoughtSpot tools?

Yes. Access is set per agent: one can get a handful of read tools, another the full set. You grant the whole MCP or single tools within it.

Where can I see what my agents did in ThoughtSpot?

Every call is logged against the agent that made it, with the tool it used. Activity can be sent on to your own monitoring.

Do I need to run the ThoughtSpot MCP server myself?

No. TrustGate connects to the ThoughtSpot MCP server and sits in front of it. Your agent only talks to your TrustGate URL, never to ThoughtSpot directly.

Connect ThoughtSpot MCP to your AI agents, securely.

Q: Is ThoughtSpot MCP secure?

The server itself exposes 13 tools, and some of them can write to your ThoughtSpot account. Behind TrustGate an agent only sees the tools you allow, requests and responses are inspected, and every call is logged. Anything outside the allowlist is denied before it reaches ThoughtSpot.

One secured endpoint for every ThoughtSpot call your AI agents make, scoped per user, inspected, and logged.

Get started for free

Get a demo

Try ThoughtSpot MCP now

What your agents can ask once ThoughtSpot MCP is connected.

Select the tools your AI agents can call

13 tools from the ThoughtSpot MCP schema.

Showing 13 tools

search_objectsreadSearches Answers, Liveboards, Liveboard visualizations and Worksheets by term, with optional filters for type, author, tag, modification date and verified status, and cursor pagination. It returns identifiers and metadata only, never the object's data, and runs no queries. ThoughtSpot notes that descriptions and query tokens are authoring metadata that can be stale and are often copied across sibling visualizations.
get_datareadFetches the full columns and rows of a saved Answer or Liveboard by re-running its saved question, with its existing filters and columns, against current data. Liveboard fetches are time-bound and return as many visualizations as fit. It is gated at runtime behind the data-download privilege.
create_analysis_sessionwriteStarts a conversational session with the Analytics Agent, optionally against a named data source or letting the Agent select the most relevant one. Annotated read-only by ThoughtSpot.
send_session_messagewriteSends a natural-language analytical question to the Analytics Agent, with an optional additional_context field carrying background the Agent should assume. The Agent then queries the warehouse on its own initiative. Annotated read-only by ThoughtSpot.
get_session_updatesreadReturns incremental updates from an analysis session, covering thinking steps, text, progress notifications and answers with an embeddable visualization URL, plus an is_done flag for polling.
create_dashboardwriteCreates a dashboard from an ordered list of answers produced in a session, with a title and a note tile of raw unescaped HTML composed by the agent, which renders for everyone who opens the dashboard.
list_orgsreadLists the Orgs the authenticated user can access with their ID, name and description, marking the one currently active. Only available when authenticated via OAuth.
switch_orghigh riskSwitches the active Org, so every subsequent tool call operates against it. ThoughtSpot states the selection is durable and shared across all of the user's sessions, persists across reconnects and applies to their other active conversations, resetting only on re-authentication or prolonged inactivity. It is annotated read-only despite this.
create_model_sessionwriteStarts a model builder session, either on a warehouse connection to build a new data model or on an existing model to edit it. ThoughtSpot instructs that the connection GUID be confirmed with the user rather than guessed or reused. Creating the session builds nothing on its own.
send_model_messagewriteSends a natural-language instruction to the model builder, or answers one of its clarifications with selected option IDs. Its description instructs that intended formulas be named to the user and confirmed before being added.
get_model_updatesreadReturns the model builder's incremental updates, including a proposed plan, clarifying choices, a task tracker, progress notifications and model state. The description marks a proposed plan as a hard stop requiring the user's explicit approval, never the agent's.
finalize_modelhigh riskSaves the built data model. Called without confirm it returns a summary for review and saves nothing; only confirm set to true persists the model and returns its GUID and URL.
check_connectivityreadVerifies that the connection and authentication are working, for diagnosing failures in other tool calls.

What ThoughtSpot MCP exposes

ThoughtSpot MCP groups its 13 tools into 6 categories, each with a suggested starting policy.

All 13 tools as JSON, ready to paste.

Category	Covers	Access	Suggested policy
Model Building	Starting a session against a warehouse connection to build a new data model or against an existing model to edit it, sending natural-language build instructions or answering the builder's clarifications, polling its updates including the proposed plan, task tracker and progress notifications, and saving the result	Read / Write	These are offered as a group, gated on the user's data-modeling privileges. finalize_model returns a summary unless confirm is set, and only confirm saves, which is a proper two-phase commit. The instructions not to auto-approve a plan or add formulas unprompted sit in tool descriptions, so they guide rather than enforce
Analysis Sessions	Creating a conversational session with the Analytics Agent against a chosen or automatically selected data source, sending a natural-language question with optional background context, and polling for incremental updates including thinking steps, text, progress notifications and answers with embeddable visualization URLs	Read / Write	ThoughtSpot annotates all three read-only, but sending a message sets the Analytics Agent querying the warehouse on its own initiative. The additional_context field carries free-text assumptions into the Agent's interpretation, so log what is passed there
Discovery & Data	Searching Answers, Liveboards, Liveboard visualizations and Worksheets by term with filters for type, author, tag, modification date and verified status, and fetching the full columns and rows of a saved Answer or Liveboard by re-running its saved question against current data	Read	The split is deliberate: search_objects returns metadata only and never contents, while get_data is gated at runtime behind the data-download privilege. get_data re-runs the saved question, so rows reflect current data rather than what was saved
Orgs	Listing the Orgs the authenticated user can access with the active one marked, and switching which Org every subsequent tool call operates against	Read / Write	switch_org is annotated read-only but changes durable state shared across the user's other active conversations and surviving reconnects. It moves which tenant's data every later call reads, so gate it and record the resulting Org id
Dashboards	Creating a dashboard from a list of answers produced in an analysis session, with a title, an ordered answer list that sets tile layout, and an HTML note tile summarising the analysis	Write	The note tile takes raw unescaped HTML that the agent composes, which then renders for everyone who opens the dashboard. Treat that field as agent-authored content published to colleagues rather than as a caption
Connectivity	A ping that verifies the connection and authentication are working, for diagnosing failures in other tool calls	Read	Harmless and worth allowing freely. It is the right first call when other tools fail, since it separates an authentication problem from a permissions one

Connect ThoughtSpot MCP to your AI agent

Four steps, under ten minutes.

01
Sign in to NeuralTrust
Open the NeuralTrust platform, sign in and go to Getting started, then Connect your tools.
02
Choose your agent
When asked which agent you use, pick Claude Code, Codex, Cursor, Windsurf, Copilot or your own code.
03
Connect ThoughtSpot
Find ThoughtSpot in the tool list and connect it.
04
Connect your agent
Follow the steps on screen to add the config to your agent, then send the test prompt.

Permissions down to the single tool

You do not allow ThoughtSpot MCP or block it. You allow the exact tools each agent can call, one by one.

ThoughtSpot MCP	13 TOOLS AVAILABLE
Code review agent	4 tools, read only
Issue triage bot	9 tools
Platform team assistant	all 13 tools

Tool by toolConnecting the server is not the unit of access. Each tool is granted on its own.
Default denyA tool you have not enabled is never exposed to the agent.
Approval on writesTools that change or delete things can require a human to confirm.

Is ThoughtSpot MCP secure?

One endpoint instead of raw credentials. Every call is checked against policy and logged.

AI agentsConnect through your MCP client configuration.	NeuralTrustApply controls based on your configuration.	ThoughtSpot MCP13 tools in the current set. ThoughtSpot ships four versioned tool sets and a deployment exposes one of them: a legacy set of five, a Spotter set of eight, that set plus four model-building tools, and finally plus get_data. Exposure is also privilege-gated at runtime, with the model tools offered as a group on data-modeling privileges and get_data on the data-download privilege. The Org tools require OAuth authentication.
Coding assistants	Connection & credentialsper user	Model Building4 tools
Internal copilots	Tool access policiesdeny by default	Analysis Sessions3 tools
Workflow automations	Request & response inspectioninline	Discovery & Data2 tools
Compatible MCP clients	Activity monitoringalerts and logs	All other tools4 tools