Overview
4DA reads the internet for developers — privately, locally. Your codebase decides what's relevant.
It scans your codebase — Cargo.toml, package.json, go.mod, Git history — and scores every article, advisory, and release from 20+ sources against what you actually build. An item needs 2 or more independent signals to survive. Everything else is rejected.
Benchmarked across 9 developer personas against a 245-item labeled corpus: 93% of content is filtered as noise, 98.9% of actual noise is correctly rejected. Measured, not asserted — see the method. Your real rejection rate — computed from your own data — is shown in the app.
Saves and dismissals build a preference profile you can inspect, pin, or forget — and teach the Brief what to stop showing you. Relevance scoring itself stays grounded in your actual stack. When the engine improves it re-judges everything it already holds: yesterday's noise becomes tomorrow's signal.
Two ways to run it
You don't have to install a desktop app to get value on day one.
| What it is | Setup | |
|---|---|---|
| MCP server | A command-line server that plugs 4DA's intelligence into Claude Code, Cursor, Windsurf, or any MCP client | npx @4da/mcp-server — no keys, no account |
| Desktop app | The full Tauri app: scored feed, briefings, radar, blind spots, Learned Preferences | Download for Windows, macOS, or Linux |
The MCP server and the desktop app read from the same local database. Start with either.
Where to go next
- Install — download the app or run the MCP server
- Quickstart — from install to your first scored feed
- The scoring engine — how an item earns its place
- Privacy & BYOK — why local-first means you don't have to trust us
All signal. No feed. 4DA is source-available under FSL-1.1-Apache-2.0 and converts to Apache 2.0 three years after each release.