LLMs.txt directory

Frequently asked questions about AgentSpot

Get answers to the most common questions about AgentSpot.

Connectors

Can we add the [X] MCP connector?

If the tool has an MCP Server, an AgentSpot admin can add it. If the tool does not have an MCP Server, Please fill a connector request here.

My Agent does not connect to <connector xyz>

Ensure that:

  1. You have signed in to the connector by doing the following:

    1. Click your user menu in the upper-right corner of AgentSpot

    2. Click Connectors.

    3. Find your connector, click Connect and sign in. your Agent has been configured to access that connector (Agent > Edit > Connector > select connector <xyz>

  2. Your Agent has been configured to access that connector, by doing the following:

  3. Open your Agent, click Edit Agent, then click the pencil icon next to Connectors.

  4. Select a connector and click Save changes.

Building agents

I gave the Agent instructions or corrections during a conversation, but they’re gone the next time I open the Agent. Why?

Instructions you type in a conversation are not saved to the Agent’s configuration. They only apply to that session, or can update the Agent or user memory.

To make a change permanent, go to Edit mode and update the Instructions there. You can tell the Agent in the builder what you want to change, and it will update the instructions for you.

If you find yourself correcting the Agent in use mode, copy that correction and paste it into Edit mode in the builder to ensure it is committed to memory.

Using agents

Can I duplicate an existing Agent to use as a starting point for my own?

Copying an Agent is not yet supported. Instead, you must open the Agent you want to base yours on, copy its instructions in the builder, create a new Agent, and paste those instructions as a starting point. The fastest way to do this, is if the Agent owner shares the instructions with you directly.

I built an Agent. How do I put it on a schedule?

Agents can’t be directly scheduled. Workflows are designed for that.

The fastest way to convert an existing Agent is the following:

  1. In a chat with your Agent, ask: "Create instructions for me to build a Workflow in AgentSpot that runs this Agent’s task on a [daily/weekly/etc.] schedule."

  2. Copy the instructions it produces.

  3. Create a new Workflow in AgentSpot and paste those instructions to guide the Workflow builder.

    you cannot invoke an existing Agent directly inside a Workflow today — the Workflow must contain its own instructions.
Can I expose my AgentSpot Agent as a Slack bot so users can interact with it without opening AgentSpot?

Yes. Here’s how to install Slack integration:

  1. Build your Agent in AgentSpot as normal.

  2. Invite AgentSpot to the public channel you’d like to interact with this Agent in. You can do that by simply typing “@AgentSpot” in that Slack channel.

  3. Ask an AgentSpot admin to bind your Agent to that specific Slack public channel. This will ensure that questions addressed to @AgentSpot in that channel are answered by the Agent of your choice.

    Users can then @AgentSpot in that channel to interact with your Agent.

    Access rules carry over. Only users who already have access to the Agent in AgentSpot can use it via Slack.

    Direct messages and private channels are currently not supporated. Slash (/) command bridges (For example: /nda-review) are also not yet supported.
How do I share an Agent with my teammates?

Open the Agent → click Share → enter your teammates' email addresses. There are no groups in AgentSpot, so you must share with each person individually. Users you share with will receive an email notification.

Workflows

When should I use an Agent, and when should I use a Workflow?

Use agents to have live conversations and iterate. Use Workflows whenever you want to schedule tasks.

You cannot invoke an existing Agent in a Workflow, but you can go to that Agent and tell it you want to create a repeatable Workflow out of what it’s doing, and then paste its instructions in the Workflow builder.

I’m having trouble with Workflows — scheduling, viewing step details, editing inputs, or the chat is greyed out.

Editing inputs directly is intentionally restricted. Each step’s output must closely lead into the next step’s input, so the Workflow builder Agent controls inputs holistically. You can change inputs by chatting with the builder.

Can users see intermediate Workflow outputs?

Not fully. Only some input/output JSON exists.

Apps

Publishing and sharing an App

Once the Agent builds an App and is shown in preview, you can choose to publish those changes so they’re saved. Note that publishing an App will only allow view access for people you’ve explicitly shared the App with. The first time you publish your App, no one else sees it until you share it.

How does access control work?

Every App inherits the access rules on your sources. ThoughtSpot Models honor your instance’s row-level security and permissions. Unstructured connectors (Salesforce, Slack, Jira, Google Workspace) authenticate per user using OAuth. Each user opening the same App sees a different, correctly-scoped version.

What kinds of blocks can be inside an App?

Apps combine data nodes (charts, KPIs, tables fetched live from your sources) and Agent nodes (LLM-generated summary blocks that interpret the data). The Agent picks which block types fit the prompt.

Which connectors are supported?

AgentSpot shares connectors across Agents, Workflows and Apps. For a complete list of supported connectors, see AgentSpot connector reference.

Can I edit an App after it’s built?

Yes. Every App can be reopened in the builder. Chat with the Agent to add sections, change styles, adjust scope, or fix issues.

Are Apps available only in AgentSpot?

Yes. Apps are only in AgentSpot, and they can connect to ThoughtSpot models.

How can I use CSV data to create Apps?

This is not possible, currently. An simple workaround is to upload it to Google Sheets and use the Google Drive connector to fetch the data.

Admin center

I’ve had AgentSpot enabled on my instance but I can’t find where to grant users access or the AgentSpot link isn’t showing in the nav.

AgentSpot roles should appear in Admin > Roles in ThoughtSpot. Assign the relevant role to your users there.

AgentSpot can’t be accessed using SSO on my instance. Do users need separate accounts?

SSO for AgentSpot is not yet supported on all instance configurations. On affected instances, users will need to sign in with basic authentication. Contact ThoughtSpot Support if you need help setting this up.

What does an Agent "user" mean in the admin usage dashboard? - Is it a person who has added the Agent to their team or had a conversation? Does having a conversation add it to a team by default?

It represents the total number of users who had a conversation with the Agent. They need not have saved it to their team to be counted there.

Guardrails and permissions

Are guardrails only instruction-based today?

Beyond instructions, admins can configure tool-level permissions per connector in staging. For example: Disabling Jira ticket creation. This is intended to prevent destructive actions caused by hallucinations.

Can permissions be controlled granularly?

Yes. Controls apply to every tool within every connection.

Can admins assign credit limits per user?

Not currently.

Auditability and privacy

What does the audit log show?

Tool-call events, such as Slack authorization or replies posted. It can be filtered by date and exported as CSV.

Does the audit log include prompts, inputs, and outputs?

Not currently. Full replay and input/output visibility are not currently possible.

Does Agent data go to the LLM?

Yes. There is no masking or filtering layer in place before data is sent to the LLM.

Is customer data used to train LLMs?

No. Enterprise-plan data is not used for LLM training.

How long are conversations retained?

One year, by default.

Are chats shared when an Agent is shared?

Not currently. Only the Agent is shared.

Billing and credits

Is billing per Agent or per conversation?

Per conversation. One continuing chat within 24 hours uses one credit.

If I have three independent chats with the same Agent, how many credits are used?

Three credits.

Spotter

Can AgentSpot use ThoughtSpot data models if the customer is still on Spotter 2?

For Agents and Workflows, Spotter 2 is supported. Apps, however, require Spotter 3 for query resolution.

Can AgentSpot send filtered Liveboard PDFs to 100 clients on a schedule?

Not currently. AgentSpot does not have Liveboard-level understanding. It works with data models.