We Used
RedisVL : To build a Knowlege base that can be used on any text box for any task. It ingest data and stores is based on topics that is smartly fetched based on context
Sanity.io: To Directly embed pointer.ai and allow it to publish and transform content on the fly without permsiions. Currently we are using it as a note taking app.
Lightpanda: We are using Lightpanda to get Real time Information on Fact Checking and current Questions
Parallel: Using as a Research Agent to fetch information on selected data to send and save information across the Web.
Postman: We connected Postman as MCP tools to show how we can integrate different tools with Google SDK and pointer.ai
đź’ˇ Inspiration
Switching between tabs and CRMs just to log a quick note or contact always broke my flow. I wanted something that could understand what I was looking at and let me take CRM actions right where I was — without copy-paste or UI overload.
⚙️ What it does
Pointer is an AI that lives on your mouse pointer. When you select text, it understands the context — names, companies, notes, or leads — and lets you instantly publish or log that data directly to your CRM. It’s like having CRM controls built into your cursor.
🧰 How we built it
I built a lightweight browser extension that captures text selections and sends them to a Go backend. The backend uses an LLM to extract structured data and push it into a simple CRM database. I instrumented the backend with Prometheus metrics and visualized performance and usage in Grafana.
🚧 Challenges we ran into
Handling text selection consistently across different web apps and inputs.
Ensuring privacy — making sure nothing sensitive gets sent without consent.
Keeping the UI subtle but responsive (floating near the cursor without blocking content).
Optimizing latency for AI extraction so actions feel instant.
🏆 Accomplishments that we’re proud of
Built an end-to-end prototype that can extract contact info and push it into a live CRM database in one click.
Integrated observability from the start using Prometheus and Grafana.
Designed a frictionless workflow that works in any tab without manual form entry.
đź“š What we learned
How to balance local client UX with backend AI logic for low-latency workflows.
How to design contextual AI interfaces that enhance existing workflows instead of replacing them.
Privacy-by-design matters early — users trust tools that are transparent about what’s being sent.
🚀 What’s next for Pointer
Expand from browser-only to desktop apps (so it works across email, docs, and chat).
Add natural-language CRM commands (“log this as a deal”, “create note for Alex”).
Deeper integrations with HubSpot, Salesforce, or custom CRMs.
Local AI summarisation for privacy and offline use.
Log in or sign up for Devpost to join the conversation.