by Width.ai

Used for keeping up with sales and customer management via an agentic framework. Integrates directly with slack.

Deals slip through the cracks. An email goes unanswered for a week. A promising lead stalls because nobody remembers the last conversation. A renewal date passes because it was buried in a CRM record no one checked. These aren't dramatic failures — they're the quiet, repeated losses that add up over a quarter.

This system exists to close that gap. It connects HubSpot, Gmail, and Google Docs behind a single natural language interface so that anyone on the team can ask a plain question and get an immediate, sourced answer:

• "Did anyone follow up with Acme Corp after the demo?" — searches emails, deals, and contact activity in seconds instead of clicking through CRM records
• "What deals are stalling in negotiation?" — surfaces deals that haven't moved stages, with context on last touch and owner
• "When was the last time we talked to Sarah at Width?" — pulls the most recent email, meeting, or note across HubSpot and Gmail
• "What proposals did we send last month?" — finds documents in Google Drive without remembering file names or folder structures

The insights pipeline goes further — it proactively scans CRM data to find patterns you didn't think to ask about: neglected contacts, deals with no recent activity, or tickets trending in a direction that needs attention.

The net result: fewer forgotten follow-ups, faster answers about deal status, and visibility into what's actually happening across your pipeline — without requiring everyone to be a CRM power user.

Search Pipeline: End-to-End Walkthrough

This document explains exactly how a natural language query flows through the system, from the initial HTTP request to the final JSON response. For architecture diagrams see architecture.md; for tool parameter reference see tools.md.

Request Intake

A query enters via POST /api/v1/query: