ThoughtSpot AgentSpot Documentation

AgentSpot is an AI-powered work assistant that connects your business data, internal knowledge, and external tools into one intelligent workspace so you can get answers, take action, and make decisions faster. AgentSpot lets your teams build and use AI agents on top of your organization’s data.

Features

Natural language chat

Ask your questions in plain English. AgentSpot understands context, remembers your conversation history, and delivers structured answers with supporting data.

ThoughtSpot analytics

Query your business data through ThoughtSpot. Get live charts, embed Liveboards directly in chat, and explore worksheets without writing a SQL query.

Multi-model AI

Choose between models from OpenAI, Anthropic, and Google, then switch models any time in your editable chats.

Deep Research

Go beyond simple answers. Deep Research mode performs multi-step investigations across the web, synthesizes findings, and delivers comprehensive reports.

Secure code execution

Run Python code, create visualizations, and process data in a secure sandbox. Results appear directly in your conversation.

Benefits

Enterprise-grade security: All code execution happens in isolated containers. Data is encrypted at rest and in transit with per-user access controls, and sensitive values are removed before output reaches the UI.

Multi-model flexibility: Plug and play any LLM — OpenAI, Anthropic, or Google. Pick the model that fits the task, then switch again whenever you want a different tradeoff.

Deep data integration: Query ThoughtSpot worksheets, embed live dashboards, process Google Drive files, and automate Gmail and Slack — all from a single chat interface.

Highly configurable: Build custom agents in Agent Studio with specific instructions, tools, and skills. Share them with your team or keep them private.

Getting started

I am a…​ My goal Start here

Administrator

Provision AgentSpot, connect identity, data, and model providers, and set governance.

For administrators

Agent builder

Build agents and workflows my team can use.

For agent builders

End user

Use an agent that someone else built.

For end users

For the full table of contents, see AgentSpot documentation.

For administrators

Work through these in order. Together they take you from a new tenant to a configured, governed AgentSpot environment.

  1. Admin Settings overview — provisioning checklist and overview of what you administer.

  2. Authentication and access — SSO, roles, groups.

  3. Connector administration — data sources, model providers, per-group sharing.

  4. Usage and limits — quotas, guardrails, audit logging.

  5. Security architecture — for your security review.

When you’re done: Your tenant is provisioned, users sign in via your IdP, the right data sources and model providers are connected, governance is set, and agent builders can start work.

For agent builders

Start with one narrow agent. Move to workflows once a single agent feels solid.

  1. Agents overview — what an agent is, when to build one.

  2. Create an agent — the full builder walkthrough.

  3. Instruction framework — how to write a good prompt.

  4. Sharing and publishing — audiences, versions, rollback.

Then, when you’re ready for automation:

  1. Workflows overview

  2. Create a workflow

  3. Example patterns — four shapes to adapt.

Before you start: Your admin must have granted you the Agent Builder role and shared at least one data source and one model provider with your group.

For end users

  1. Using a published agent — find an agent, ask good questions, read answers, know when to trust them.

Before you start: Your admin must have given you access to at least one agent.

Common questions

Do I need to know how to code?

No. Administration, agent building, and agent use are all no-code.

How is AgentSpot different from a chatbot?

A chatbot answers from general knowledge. An AgentSpot agent answers from your data and tools, under controls your organization sets. Each agent is scoped to a specific job, with explicit data sources and explicit limits.

Is my data sent to a third-party model provider?

It depends on which model providers your admin has connected and which model an agent uses. Your admin can answer this for your tenant.

Where do I report a problem?

  • Wrong answer or missing data on a specific agent: contact the Owner on the agent’s detail page.

  • Can’t sign in, can’t see any agents: contact your AgentSpot admin.

  • Suspected security issue: contact your admin immediately.

Next steps

  • Administrators: schedule a 30-day review of usage, audit trends, and quota headroom.

  • Builders: publish your first agent to a small group, gather feedback for a week, then iterate.

  • End users: try the agents in your Agents list and use thumbs-up / thumbs-down to send feedback.

Was this page helpful?
×