DeepPulse AI v2
An experimental multi-agent platform for business research, coordinated through a real-time dashboard.
- Status
- Experimental
- Year
- 2026
- Role
- Design and development
- Links
- Open the demo (opens in a new tab) Source code is private
Live

Overview
DeepPulse explores what a small team of AI agents can do for someone planning a product launch: look for product opportunities, analyse competitors, draft ad copy, model margins and risk, and track launch work. The agents report into a live dashboard so you can watch the work happen rather than wait for a single answer.
What I built
- A monorepo with npm workspaces: a React and Vite dashboard, an Express API with a WebSocket event hub, and a background worker for agent tasks.
- An agent registry with five starter agents, plus shared packages for financial models (margins, break-even, risk), security policies and event definitions.
- Abuse protection on the API: rate limiting, signup caps that survive restarts, per-IP limits on expensive endpoints, and Helmet headers.
- Plan limits wired to Stripe in test mode, and a Product Finder that compares listings across marketplaces and scores arbitrage opportunities.
Honest limits
- Experimental, not a production service. The data scrapers are mocked, so results are not based on live market data.
- Payments run in Stripe test mode only. The repository is private.
Screenshots


