timeahead-mcp

MCP server for timeahead.in — live, independent trust scores for MCP servers.

timeahead scores 800+ MCP servers nightly across five dimensions: security (35%), freshness (25%), adoption (20%), quality (10%), and trust (10%). This package gives any AI assistant (Claude Desktop, Cursor, Cline, Continue) direct access to those scores so it can answer questions like "which Postgres MCP is safest?" with live data instead of stale training.

Install

# One-time global install
uvx timeahead-mcp

# Or add to mcp.json
{
  "mcpServers": {
    "timeahead": {
      "command": "uvx",
      "args": ["timeahead-mcp"],
      "env": {
        "TIMEAHEAD_API_KEY": "ta_mcp_..."
      }
    }
  }
}

The TIMEAHEAD_API_KEY is optional. Without it you get 200 requests/hour (anonymous). Get a free key at timeahead.in for 500 req/hr.

Tools

Discovery

Trust & Inspection

Risk & History

Recommendations

AI News Feed

Example prompts

"Which Postgres MCP server should I install?"
→ search_servers("postgres", min_score=70)

"Is my current MCP setup safe? Here are my installed servers: [slugs]"
→ audit_my_stack([...])

"What changed in MCP security this week?"
→ get_score_drops(days=7)

"Find me a safer alternative to filesystem-mcp"
→ recommend_safer_alternative("filesystem-mcp")

"Should I use mcp-server-postgres or mcp-postgres-pro?"
→ compare_servers(["mcp-server-postgres", "mcp-postgres-pro"])

"What's happening in AI this week?"
→ get_latest_news(limit=5)

Environment variables

Scoring methodology

Scores are computed nightly by scanning each server's repository (Gitleaks for secrets, AST analysis for capability flags), fetching GitHub maintenance metrics, and checking weekly download counts from npm/PyPI.

timeahead's own MCP server is scored on the same rubric with no special treatment. View the methodology.

License

MIT — see LICENSE.