RepoLens

Use from your agent

The local RepoLens MCP server lets Claude, Cursor, and other agents vet a dependency before adding it.

Coding agents pick packages from README text, star counts, and stale blog posts. The RepoLens MCP server gives them a due-diligence step instead: the agent gets structured JSON, and RepoLens writes a self-contained HTML report and opens it in your browser so you see the same verdict.

It runs locally over stdio, separate from the extension. Nothing is sent to RepoLens — see Privacy.

Tools

  • scan_repo — verdict-first report: fit, health, pros, cons, red flags, capabilities, bottom line.
  • deep_dive — plain-English architecture explanation with weak spots and assumptions.
  • blueprint_scene — graph-shaped architecture map.
  • compare_repos — 2–5 repos or packages for one use case: a default pick, ranking, and tradeoff matrix.
  • evaluate_for_goal — one repo against a concrete goal and constraints: adopt / trial / hold / reject, blockers, and a trial plan.
  • analyze_product — a public product page: its model, core loop, dependencies, and which claims still need evidence.

scan_repo and compare_repos accept GitHub, GitLab, npm, and PyPI inputs.

Set it up

From a checkout of the repo:

cd mcp
npm install
export ANTHROPIC_API_KEY=sk-ant-...   # or OPENAI_API_KEY, OPENROUTER_API_KEY, GOOGLE_API_KEY
export GITHUB_TOKEN=ghp_...           # optional; lifts GitHub's 60 requests/hour limit
node server.js

For Claude Desktop, add it to claude_desktop_config.json:

{
  "mcpServers": {
    "repolens": {
      "command": "node",
      "args": ["/absolute/path/to/repolens/mcp/server.js"],
      "env": { "ANTHROPIC_API_KEY": "sk-ant-..." }
    }
  }
}

Every environment option is in mcp/README.md.

Ask for it

Use RepoLens to check whether I should use honojs/hono for an edge API.

Before installing this package, run a RepoLens scan and open the report.

Compare honojs/hono vs fastify/fastify for an edge API and open the report.

RepoLens tells your agent whether to depend on a piece of code. When what the agent needs is a website without its own MCP server, wmcp.sh turns the site's URL into agent-callable tools — and analyze_product can inspect that product before you wire it in.