How I Built an AI Powered OSS Triage Dashboard Using Coral and Claude in One Day

Every morning, open source maintainers wake up to hundreds of GitHub issues and PRs. They do not know what is urgent, what is stale, what is a duplicate. It kills hours of productive time. I built OSS First Mate to fix that.

This is the story of how I built it in one day for the Pirates of the Coral-Bean Hackathon by WeMakeDevs, and exactly how you can build something similar.

What is OSS First Mate

OSS First Mate is an AI powered triage dashboard for open source maintainers. You enter any GitHub repo and it instantly surfaces what needs attention today. It queries GitHub, Slack and Notion simultaneously using Coral SQL, passes the combined data to Claude AI and returns a repo health score, stale PR list, duplicate issue clusters and a top priority action for the day.

Live demo: https://coral-pro-frontend.vercel.app GitHub: https://github.com/alamzia23/oss-first-mate Demo video: https://www.youtube.com/watch?v=EfmFvH7Ypmg

The Tech Stack

Frontend: Next.js with Tailwind CSS Backend: Node.js with Express Data Layer: Coral for cross source SQL queries AI Layer: Claude AI via Anthropic API Deployment: Vercel for the frontend

What is Coral and Why It Matters

Coral is a tool that gives you a single SQL interface to query multiple data sources simultaneously. Instead of writing separate API clients for GitHub, Slack and Notion, you write one SQL query and Coral handles everything.

Without Coral this project would have needed three separate API integrations, three different authentication flows and manual data merging logic in the backend. Coral collapsed all of that into a single SQL interface.

Here is what a Coral query looks like in practice:

coral sql — format json “SELECT number, title, body, created_at, comments, user__login FROM github.issues WHERE owner=’facebook’ AND repo=’react’ AND state=’open’ LIMIT 30”

That single command fetches live data from GitHub with no API key setup, no pagination logic and no response parsing needed.

How I Built It :-

Step 1: Install and Set Up Coral

npm install -g @withcoral/cli coral source add github coral source add slack

After running these commands, Coral authenticates with GitHub and Slack through its own auth flow. You never need to manage tokens manually.

Step 2: Build the Backend

The backend is a simple Express server with one main endpoint. When a user enters a GitHub repo, the server runs three Coral SQL queries in parallel. One query fetches open issues from GitHub. One query fetches open PRs from GitHub. One query fetches recent Slack messages from the connected workspace.

The results are combined into a single JSON response.

Step 3: Add Claude AI Insights

The combined Coral data is sent to Claude AI via the Anthropic API. Claude analyzes the data and returns a structured response containing a repo health score out of 100, a top priority action for today, a pattern detection summary and risk flags.

Step 4: Build the Frontend

The frontend is a Next.js app with five tabs. The Urgent tab shows issues with high comment counts. The AI Slop tab shows low quality or AI generated issues. The Stale tab shows PRs older than 7 days. The Duplicates tab shows clustered similar issues. The Slack tab shows recent team messages.

What I Learned

Coral is genuinely powerful for hackathon projects. The ability to JOIN data across GitHub, Slack and Notion in a single SQL query is something that would normally take days to build from scratch. I built a working multi source AI agent in one day because Coral handled all the data plumbing.

Claude AI is excellent at reasoning about structured data. Passing the raw Coral output directly to Claude and asking it to prioritize gives surprisingly good results with minimal prompt engineering.

Building for a real use case matters. OSS First Mate solves a problem I genuinely experience as a developer. That focus made every decision easier.

What You Can Build With This Approach

The same pattern of Coral for data plus Claude for reasoning works for many other use cases. A customer support dashboard that joins Slack, Notion and GitHub issues. A personal productivity agent that joins Google Calendar, Gmail and Notion. A sales intelligence tool that joins CRM data, email and Slack conversations.

Coral makes the data layer trivial. Claude makes the reasoning layer trivial. You just need the idea.

Try It Yourself

GitHub: https://github.com/alamzia23/oss-first-mate Live: https://coral-pro-frontend.vercel.app Demo: https://www.youtube.com/watch?v=EfmFvH7Ypmg

If you are building something with Coral I would love to see it. Drop a comment or find me on LinkedIn.

Built for the Pirates of the Coral-Bean Hackathon by WeMakeDevs.