@4da/mcp-server

smithery badge

Dependency intelligence for AI coding agents. Live CVE scanning, dependency health checks, upgrade planning, ecosystem news, and persistent decision memory. Zero config, privacy-first.

You:     "Check my dependency health"
Claude:  Health: 72/100. 47 dependencies scanned, 3 vulnerable, 1 deprecated, 8 outdated.

         CRITICAL  openssl-sys 0.9.93  CVE-2025-4231        -> 0.9.96
         HIGH      serde       1.0.197 RUSTSEC-2026-12      -> 1.0.210
         MEDIUM    cookie      0.17.0  deprecated           -> 0.18.1

         Quick wins: 6 patch upgrades, 2 minor. Run upgrade_planner for full plan.

One command to install. No API keys. No accounts. No code leaves your machine.

Install

claude mcp add 4da -- npx @4da/mcp-server

{
  "mcpServers": {
    "4da": {
      "command": "npx",
      "args": ["@4da/mcp-server"]
    }
  }
}

{
  "servers": {
    "4da": {
      "type": "stdio",
      "command": "npx",
      "args": ["@4da/mcp-server"]
    }
  }
}

{
  "mcpServers": {
    "4da": {
      "command": "npx",
      "args": ["@4da/mcp-server"]
    }
  }
}

npx @4da/mcp-server --setup

Then ask your AI: "Check my dependency health" or "Scan for vulnerabilities"

How It Works

On startup, the server reads your manifest and lock files (package.json, Cargo.toml, go.mod, pyproject.toml), resolves exact dependency versions, and queries live APIs. It re-reads them whenever a lockfile changes, so a long-running server never answers for yesterday's dependency set. For npm it also checks what node_modules actually holds: an installed copy that differs from the lockfile is reported with its reinstall command, instead of hiding behind a patched lockfile.

• OSV.dev for known CVEs across all ecosystems
• npm registry for version freshness, deprecation status, and weekly downloads
• crates.io sparse index for Rust package versions (avoids the 1 req/s API limit)
• PyPI JSON API for Python package metadata with license normalization
• Go module proxy for Go module versions
• Hacker News Algolia API for ecosystem news filtered by your tech stack

Results are cached (24h for registry data, 1h for vulnerabilities, 30min for news) and rate-limited per source.

What's sent over the network: package names + versions, generic tech keywords. The same data visible in your package.json. No source code, no file paths, no personal data. Set FOURDA_OFFLINE=true to disable all network calls.

The one exception: if you explicitly configure an OpenAI embedding provider (FOURDA_EMBED_PROVIDER=openai) for semantic recall, the decision/memory text you store is sent to OpenAI to be embedded. The default — no embedding provider, or a local Ollama one — keeps everything on your machine, and FOURDA_OFFLINE=true overrides it regardless.

Ecosystems supported: npm, crates.io (Rust), PyPI (Python), Go.

What You Can Ask

"Check my dependency health"                  -> dependency_health
"Scan for vulnerabilities"                    -> vulnerability_scan
"Which deps should I upgrade first?"          -> upgrade_planner
"What should I know before I start coding?"   -> what_should_i_know
"What's happening in the ecosystem?"          -> ecosystem_pulse
"What's my tech stack?"                       -> get_context
"Record a decision: we chose Postgres"        -> decision_memory
"Does switching to MySQL align?"              -> check_decision_alignment
"Remember: never use ORM for batch inserts"   -> agent_memory

All 14 Tools

Dependency Security

Intelligence

Decisions & Memory

Identity

* Requires the 4DA desktop app for full data.

Standalone vs. Full Mode

The MCP server works without the desktop app. On first run it creates a local database and scans your project:

Download 4DA for the full experience.

Transports

stdio (default) -- works with all MCP hosts:

npx @4da/mcp-server

Streamable HTTP -- for remote or multi-client setups:

npx @4da/mcp-server --http --port 4840

The HTTP transport binds to 127.0.0.1 by default and applies a Host-header DNS rebinding guard to every request. Exposing it beyond this machine requires a shared secret:

MCP_AUTH_SECRET=<same value as the relay's JWT_SECRET> \
MCP_ALLOWED_HOSTS=mcp.internal \
  npx @4da/mcp-server --http --host 0.0.0.0

Without MCP_AUTH_SECRET a non-loopback bind is refused at startup. With it, every request must carry a Bearer token whose HMAC-SHA256 signature verifies against that secret, and the token's role is enforced per tool (viewer is read-only; member and admin may write). Put TLS in front of it.

CLI Reference

npx @4da/mcp-server              # Start server (stdio)
npx @4da/mcp-server --http       # Start server (Streamable HTTP)
npx @4da/mcp-server --setup      # Auto-configure your editors
npx @4da/mcp-server --doctor     # Verify installation health
npx @4da/mcp-server --version    # Print version

Environment Variables

FAQ

Does this send my code anywhere? No. The server sends package names and versions to public APIs (OSV.dev, npm registry, crates.io, PyPI, Go proxy) and generic tech keywords to HN Algolia. The same public data visible in your package.json. No source code, no file paths, no personal data. Set FOURDA_OFFLINE=true to disable all network calls. (The sole exception is opt-in OpenAI embeddings — see the network note above.)

Do I need the 4DA desktop app? No. 9 tools work standalone: vulnerability scanning, dependency health, upgrade planning, ecosystem news, pre-task briefings, project context, decision memory, alignment checking, and agent memory. The desktop app adds a scored content feed from 20+ sources, graded against your actual stack.

Which AI tools does this work with? Any tool that supports MCP: Claude Code, Claude Desktop, Cursor, Windsurf, VS Code (Copilot), and any custom MCP client.

Build from Source

git clone https://github.com/4DA-Systems/4DA.git
cd 4DA/mcp-4da-server
pnpm install
pnpm build
pnpm test    # 71 contract tests

License

Apache License 2.0 (Apache-2.0). See LICENSE.

Built by 4DA