4da Mcp Server

What is 4da Mcp Server?

4da Mcp Server is an MCP-compatible integration that scans your project, detects your stack, and provides live vulnerability scanning, dependency health, upgrade planning, and ecosystem intelligence to your AI assistant. It works standalone with no API keys or accounts required and runs locally.

How to use 4da Mcp Server?

For Claude Code, Cursor, or Windsurf, run npx @4da/mcp-server in your project directory. This scans your project and gives your AI assistant the intelligence tools. 9 tools work standalone with zero setup; 5 more require additional configuration.

Key features of 4da Mcp Server

• 14 tools for dependency security, intelligence, decisions, and agent memory
• Scans Cargo.toml, package.json, go.mod, and Git history
• Scores articles from 20+ sources against your actual build
• 92% noise rejection rate; 98% of actual noise correctly rejected
• Behavior learning from save/dismiss feedback
• Fully local-first with no 4DA-operated server

Use cases of 4da Mcp Server

• Live vulnerability scanning for your project’s dependencies
• Dependency freshness and health monitoring
• Upgrade planning and prioritization
• Pre-task briefings with tailored ecosystem intelligence
• Cross-session, cross-agent persistent decision memory

FAQ from 4da Mcp Server

How does 4da Mcp Server differ from the desktop app?

The MCP server works standalone—no desktop app required—and gives your AI assistant live scanning and intelligence directly in your terminal or editor.

What dependencies or runtime does it require?

It runs via npx from the npm package @4da/mcp-server. The MCP server itself is a Node.js tool; it does not require a Rust toolchain. For LLM verification, you can use Ollama (free, local) or bring your own Anthropic/OpenAI key.

Where does my data live?

All indexed content, scores, and intelligence live in a SQLite database on your machine. There is no 4DA-operated server, no analytics, and no user account system.

What transport or authentication does the MCP server use?

The MCP server uses standard MCP transport. No API keys or accounts are required for the core functionality; optional LLM features require you to bring your own API key for Anthropic/OpenAI or use local Ollama.

Are there any known limits?

The MCP server provides 14 tools total: 9 work with zero setup, 5 require additional configuration. The noise rejection algorithm requires your local file system to compute accurate scores.