query-agent
Purpose: Run a thematic discovery analysis across the customer’s ingested feedback to identify and quantify topics.
Description (as exposed to the AI client): Ask Bagel’s Explore a product or customer question. Bagel ingests and synthesizes product insights across your customer feedback (support tickets, sales/customer calls, surveys, RFEs, Slack channels, CRM notes, customer interviews in-product feedback, etc.) and based on the query, returns a structured analysis: themed topics, each with a topic_id, supporting customer count, ARR impact, and a short summary. Use this for questions like “what are the most common requests/pain points related to capability/topic?” or “what are the top needs in a specific customer segment / timeframe?”. For subsequent queries, use whenever the user references any topic or finding from a prior response — phrasings like “drill down”, “tell me more about [topic]”, or even just naming a topic by title. Sessions are tracked automatically; you do not need to pass session_id.
Access: Read-only. No data is created, modified, or deleted.
Parameters:
Name | Type | Required | Description |
message | string | Yes | The product or business question to analyze (e.g. “What are the most common feature requests in the reporting area?”). |
session_id | string | No | Session identifier. Omit for most calls — the session is resolved automatically from the connection. Pass explicitly only to resume a prior session after reconnecting. |
get-evidence
Purpose: Retrieve the source records (customer quotes, call excerpts, ticket text) behind a topic surfaced by a prior query-agent analysis.
Description (as exposed to the AI client): Fetch the full underlying evidence records behind a topic returned by a prior query-agent analysis. (evidence records include textual content such as requests and pain points, customer quotes, related customer attributes (e.g. name, tier, status, ARR), Evidence attributes (e.g. priority, status, submitter, creation date, assigned product idea etc.). Do NOT use this for queries with the intent to drill into a topic that was returned by a previous query-agent call - use query-agent for that. This tool should only be used when the user shows a clear intent to use the actual evidence records for specific use cases such as exporting them to CSV, create a feature brief from an uploaded template based on the evidence records, marry the data of the evidence records with other data etc. Pass the topic_id (24-char Mongo ObjectId) from the prior response; session is resolved automatically.
Access: Read-only. No data is created, modified, or deleted.
Parameters:
Name | Type | Required | Description |
topic_id | string | Yes (unless evidence_ids is supplied) | The topic identifier (24-character hex) returned by a prior query-agent response. |
evidence_ids | array of string | No | Specific evidence record IDs to fetch. Overrides topic_id when supplied. |
page | number | No | Page number, 0-based. Default 0. |
page_size | number | No | Items per page. Default 10, maximum 25. |
session_id | string | No | Session identifier. Ignored when topic_id is provided (session is resolved from the artifact). Only relevant as context when passing raw evidence_ids without a topic. |