The facts, with their sources

What it does

Writes code: PartlySpotterCode generates embedded-analytics integration code in the IDE, but only ThoughtSpot SDK, auth and theming work rather than general application code.Skip weeks of SDK ramp-up. SpotterCode writes the integration, configures auth, and sets up your cluster without leaving the editor.thoughtspot.com/product/agents/spotterc…
Debugs and fixes: PartlySpotterCode is described as helping developers troubleshoot embedded analytics as they code; the vendor never claims it diagnoses a failure and lands the fix itself.SpotterCode helps you generate, configure, and troubleshoot embedded analytics as you code.thoughtspot.com/product/agents/spotterc…
Whole-codebase work: PartlySpotterCode reads the surrounding project structure to keep suggestions consistent, but the vendor claims no repository-wide migration or multi-file refactor capability.Suggestions read your project structure and follow current SDK conventions, so implementations stay maintainable as adoption grows.thoughtspot.com/product/agents/spotterc…
Ships it: PartlyThe coding agent executes REST and cluster operations from the editor, yet every code change it proposes waits for a developer to approve it rather than shipping itself.Your team reviews, refines, and approves every change SpotterCode proposes.thoughtspot.com/product/agents/spotterc…
Done-for-you service: NoThoughtSpot sells software a customer runs itself: self-serve Essentials and Pro plans with free trials, not a team doing client work for you.Unified platform to power trusted insights across analytics, applications, and workflows.thoughtspot.com

How much it does unattended

Runs autonomously: PartlyRead-only analysis runs without a human, but any agent action that writes to or deletes in another application is held for user approval.Read-only tools run freely. Anything that writes to or deletes from another app waits for a user to approve it.thoughtspot.com/product/agents
Multi-agent: PartlyFour named agents cover different stages of the analytics workflow on shared semantics, but the vendor does not claim they coordinate on a single task.Every agent runs on the same semantic layer and the same security model, so speed never costs you control.thoughtspot.com/product/agents
Agent permissions: PartlyAgents inherit role-based, row- and column-level security with no extra setup, and writes elsewhere need approval, so autonomy is bounded by existing data permissions.Role-based access plus row-level and column-level security apply to every agent, with no new rules to set up.thoughtspot.com/product/agents

Which models it runs on

Claude: YesAnthropic Claude is among the LLMs the platform uses, reached through Google Vertex AI rather than contracted directly by the buyer.Google Vertex AI LLMs, including Gemini and Anthropic Claudethoughtspot.com/trust/enterprise-grade-…
GPT: YesGPT models are used via Microsoft Azure OpenAI, listed by the vendor as part of the platform's disclosed model set.Microsoft Azure OpenAI GPT LLMsthoughtspot.com/trust/enterprise-grade-…
Open models: PartlyOpen-weight Mistral is reachable only through Snowflake Cortex AI; the vendor documents no way to point the platform at a model you host yourself.Snowflake Cortex AI LLMs, including Mistralthoughtspot.com/trust/enterprise-grade-…
Model choice: PartlyYou can bring your own model through five named clouds, but the plan table lists 'Flexible LLM selection' only against Enterprise, so entry tiers take managed models.Use ThoughtSpot-managed models, or bring your own through Bedrock, Azure OpenAI, Databricks, Vertex AI, or Snowflake Cortex.thoughtspot.com/product/agents

Where you use it

In your editor: YesSpotterCode is installed as an MCP server in AI-native IDEs including Cursor and Visual Studio Code, working on your open project.This guide walks you through the process of adding SpotterCode to your AI-native IDE.developers.thoughtspot.com/docs/integra…
On the command line: PartlyThere is no ThoughtSpot CLI of its own; the terminal route is registering the SpotterCode MCP server inside the Claude Code command line.For the Claude Code-only setup, you must manually point Claude Code to the SpotterCode MCP server via CLI.developers.thoughtspot.com/docs/integra…
In your pipeline: PartlyBearer-token auth exists specifically for CI/CD use, but the vendor claims no pull-request review or pipeline action by the agent itself.For programmatic or CI/CD scenarios where an MCP client cannot perform an interactive login, SpotterCode accepts a bearer token directlydevelopers.thoughtspot.com/docs/integra…
In a browser: YesThe core product is a hosted web application, ThoughtSpot Cloud, with Liveboards, search and Spotter chat in the browser plus iOS and Android apps.ThoughtSpot Cloud is AI-powered analytics that puts self-service analysis in the hands of everyone in your organization—no SQL or data expertise…docs.thoughtspot.com/cloud/26.10.0.cl

What it costs to run

How it meters: YesTwo published models: per user per month on Essentials and Pro, or consumption priced per credit if you pick the Usage pricing option.Starting as low as $0.10 Per creditthoughtspot.com/pricing
Free tier: PartlyYou can start without a sales call on a 14-day trial with no card, and the Embedded Developer plan is free for a year, but no plan stays free afterwards.Start your 14-day free trial No credit card requiredthoughtspot.com/trial
API access: YesA documented REST API v2.0 covers metadata, data, users, orgs and audit logs, with TypeScript, Java, C# and Python SDKs; embedding for internal apps is an add-on on Analytics plans.To export the TML representation of the metadata objects from ThoughtSpot in JSON or YAML format, use the following endpointdevelopers.thoughtspot.com/docs/tml-exp…
MCP server: PartlyA hosted Spotter MCP server lets Claude, ChatGPT or a custom client query governed data, but it is sold as an add-on limited to Enterprise and Embedded licences.ThoughtSpot Spotter MCP Server is available as an add-on with the following license plans: ThoughtSpot Enterprise Edition, ThoughtSpot Embeddeddevelopers.thoughtspot.com/docs/mcp-int…
Bring your own key: PartlyYou may connect your own governed model through Bedrock, Azure OpenAI, Databricks, Vertex AI or Snowflake Cortex, and then that provider bills you separately.Connect the model you already trust and govern. Conversation monitoring shows what was asked, what was answered, and how it was produced.thoughtspot.com/product/agents/spotter

Buying it for a team

A company can buy it: YesPlans are sold to organisations with user floors and ceilings, multi-tenant Orgs and an Enterprise tier quoted per deployment rather than to individuals.Custom pricing tailored to your teams, data, and deployment needs.thoughtspot.com/pricing
Seat model: YesSeats carry both a floor and a ceiling: Essentials runs from 5 to 50 users, Pro up to 1000 users, and only Enterprise is unlimited.thoughtspot.com/pricing
Pooled budget: PartlyUnder user pricing each seat gets its own Spotter allowance of 25 queries a month; the alternative Usage model bills a shared credit balance instead.Spotter AI Agents (25 queries/month per user)thoughtspot.com/pricing
Admin controls: YesAdministrators switch AI features on selectively with granular permissions, and data security rules at object, column and row level constrain what agents can reach.You enable AI features in the ThoughtSpot Agentic Analytics Platform as desired, using granular permissions and controls ensuring AI is used the way…thoughtspot.com/trust/enterprise-grade-…
Audit log: YesSecurity audit logs record logins, user and Org changes and CRUD operations, and can be pulled by REST API or pushed to your own SIEM.ThoughtSpot Cloud logs events related to user activities, account authentication, and CRUD operations.developers.thoughtspot.com/docs/audit-l…
Single sign-on: PartlySAML, OAuth and OIDC sign-on is supported, but the plan comparison shows it checked only against Pro and Enterprise, not the entry Essentials tier.ThoughtSpot supports multi-factored authentication and single-sign on via SAML and OIDC compliance identity providers.thoughtspot.com/trust/security

What happens to your code

Opt out of training: YesNo opt-out is needed: the vendor states customer data is never retained by the LLM or used for model training, and provider models delete data after processing.Your data is never retained by the LLM or used to train models.thoughtspot.com/product/agents
Data residency: YesCustomers choose a residency region for the hosted deployment, with single-tenant and customer-managed key options named for Spotter.ThoughtSpot understands the importance of data residency and offers options to select data residency regions based on your organization's needs and…thoughtspot.com/trust/security
Getting out: PartlyContent and models export as TML through documented REST endpoints and data is removed on termination, but nothing is published about expiry of unused usage credits.If you decide to no longer use ThoughtSpot, we ensure a secure and complete removal of all your data and your ThoughtSpot environment from our…thoughtspot.com/trust/security
Certifications: YesThe vendor lists SOC 1, SOC 2, SOC 3, ISO 27001 and CSA Star Level 1 certifications, with GDPR, HIPAA and DPF commitments on the Trust Center.We adhere to rigorous industry standards which are demonstrated through the security controls and compliance certifications we maintain, including:…thoughtspot.com/trust/security

Plans and pricing

Price and terms

Free to start: PartlyPartly14-day free trial with no credit card required
The terms in full
the ThoughtSpot Embedded Developer plan is Free for 1 year, up to 10 users and 25M rows of data. No standing $0 plan after either period.