# MCP Server with Spotter 3 capabilities > For the complete documentation index, see [llms.txt](https://developers.thoughtspot.com/docs/llms.txt) Source: https://developers.thoughtspot.com/docs/mcp-server-spotter3 # MCP Server with Spotter 3 capabilities ThoughtSpot Spotter MCP Server supports [Spotter 3](https://docs.thoughtspot.com/cloud/latest/spotter-versions), which enables advanced analytics, forecasting, multi-step reasoning and analysis, and deep research capabilities. The MCP Server with Spotter 3 introduces new MCP tools, real-time streaming responses, session-based conversations, and richer interactions. | | Legacy MCP Server setup | MCP Server with Spotter 3 | | --- | --- | --- | | **Functional differences** | Limited capabilities for complex analysis and context integration. | - Supports advanced analytics, forecasting, and deep research capabilities. - Supports asynchronous streaming, session-based conversations, and context integration. | | **Recommended for** | Maintaining existing integrations only. | All new integrations | | **Integration pattern Session model Response processing** | Synchronous and stateless integration pattern. Each tool call is independent. Does not support persistent sessions and requires prior context to be injected manually on every follow-up call. | Asynchronous and stateful sessions. Allows multiple follow-up questions within a session. Responses are streamed with intermediate updates, including thinking narration and clarifications. Context is retained automatically across all follow-up questions within the session. | | **Data source selection** | Requires a separate tool call (`getDataSourceSuggestions`) to retrieve source suggestions. | Built into session creation. If the `data_source_id` is not explicitly specified in the query, the most relevant source is automatically selected based on the user’s query. | | **Response delivery** | Returns full response in a single synchronous call. | Responses are streamed incrementally; updates are accumulated via polling. | | **Follow-up questions** | Requires prior context on every follow-up call. | The session retains context automatically. Optionally, the `additional_context` parameter can be included to inject new external context for each follow-up question or message. | | **API versioning support** | Not available. | Supports date-based API versioning, which is identified in the MCP Server URL as `?api-version=YYYY-MM-DD`. | The following figure illustrates the MCP architecture, tool calls, and workflow in the new MCP Server version: ![MCP architecture new](/docs/doc-images/images/mcp-architecture-new.png) ## Tool calls and workflow processing The workflow in the MCP Server with Spotter 3 typically includes the following stages: 1. **User asks a question** A user sends a query in the chat interface to get data insights. For example, _What were the total sales of Jackets and Bags in the Northeast last year?_ 2. **Agent calls `create_analysis_session`** The agent calls `create_analysis_session` to establish a stateful analytical session with ThoughtSpot’s Spotter 3 engine. ThoughtSpot returns an `analytical_session_id` that identifies the session for all subsequent calls. If the user specified a data source, the agent passes it as the optional `data_source_id` parameter. If not specified, ThoughtSpot automatically selects the most relevant data source based on the question. 3. **Agent calls `send_session_message`** The agent calls `send_session_message` with the `analytical_session_id` and the user’s question. Optionally, the agent can pass `additional_context` to provide external information that should influence the analysis. For example, `The user’s fiscal year starts in April` or `The user is a manager of the West region`. ThoughtSpot confirms receipt with a `success` response and begins processing the question asynchronously. The answer is not returned in this call. 4. **Agent polls `get_session_updates` for results** The agent calls `get_session_updates` repeatedly in a loop until `is_done` is `true`. ThoughtSpot streams updates incrementally across multiple polls. The agent must accumulate updates from every poll. 1. **User generates a dashboard with the results (optional)** The user can choose to save answers from the conversation as a ThoughtSpot dashboard. For this workflow, the agent extracts the `answer_id` and `answer_title` from each `answer` type update collected during polling and calls `create_dashboard`. ThoughtSpot creates a dashboard and returns a `dashboard_id` and a `dashboard_url`, a direct link to the dashboard in the ThoughtSpot UI. The other optional tool calls include: - **User asks a follow-up question (optional)** The user can ask a follow-up question in the same session. The agent calls `send_session_message` again using the same `analytical_session_id`. ThoughtSpot retains the full conversation context automatically. The agent returns to step 4 to poll for the follow-up response. - **User switches to a different Org (optional)** If the user belongs to more than one Org and connects over OAuth, the agent calls `list_orgs` to return all accessible Orgs and flags the Org that the user is currently logged in. When the user requests a switch, the agent calls `switch_org` with the target `org_id`. For more information, see [Org switching tools]({{navprefix}}/{{mcp-tool-reference-spotter3}}#org-switching-tools). - **Agent searches for existing content (optional)** If the user refers to content that already exists, for example, "open the regional sales Liveboard", the agent calls `search_objects` with a search term and optional filters such as object type, author, tag, or verified status. `search_objects` returns a ranked list of matching objects with their metadata and deep links. ## Additional resources - For information about MCP, see the [Model Context Protocol specification](https://modelcontextprotocol.io). - For implementation details, see the [MCP Server GitHub repository](https://github.com/thoughtspot/mcp-server).