Analook MCP Server

Use Analook from Claude Desktop, Cursor, or any MCP-compatible client — no installs, no local processes. One HTTPS endpoint, one Bearer token, five tools for running and reading competitor analyses from inside your agent workflow.

Endpoint: https://analook.com/mcp Transport: Streamable HTTP Protocol: MCP 2024-11-05

🖥️ Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows) and add:

{
  "mcpServers": {
    "analook": {
      "url": "https://analook.com/mcp",
      "headers": {
        "Authorization": "Bearer <YOUR_ANALOOK_TOKEN>"
      }
    }
  }
}

Restart Claude Desktop after editing. Then ask Claude: "use Analook to analyze linear.app".

⌨️ Cursor

Open Settings → MCP → Add new MCP server. Use the same JSON as above, or manually:

Name:     analook
URL:      https://analook.com/mcp
Headers:  Authorization: Bearer <YOUR_ANALOOK_TOKEN>

🔑 Getting your token

Sign in to analook.com
Open your browser DevTools (⌘⌥I / F12) → Application → Local Storage → analook.com
Copy the value of sb-*-auth-token → the access_token field
Paste into the Bearer <...> slot above

Tokens expire in ~1 hour by default. A dedicated long-lived API key endpoint is on the roadmap — ping us if you need it sooner.

🛠️ Available tools

analyze_competitor(url, product_name?) ⚡ 1 credit · auth required

Kick off a full analysis across 15+ data sources. Returns job_id.

Example: analyze_competitor(url="lovable.dev")

get_report_status(job_id) public

Poll until status="completed". Typical run is 2–5 min.

get_report(job_id) public

Full structured JSON report — website, SEO/traffic, social, Product Hunt, GitHub, pricing, funding, AI insights, growth playbooks.

get_report_markdown(job_id) public

Human-readable Markdown — pipe directly to agents or paste into docs.

list_my_reports() auth required

List your last 50 reports (id, url, product_name, created_at, status).

💡 Example agent workflow

Ask your agent: "Analyze my top 3 competitors and tell me where I'm losing."

1. analyze_competitor(url="linear.app")        → job_id: "abc123"
2. analyze_competitor(url="notion.so")         → job_id: "def456"
3. analyze_competitor(url="asana.com")         → job_id: "ghi789"
4. [wait 3 min]
5. get_report("abc123"), get_report("def456"), get_report("ghi789")
6. Agent synthesizes: "Linear leads on DX (open-source energy + GitHub stars),
   Notion on content (blog traffic 3x yours), Asana on enterprise (pricing tier
   gap). Your weak spot: zero Product Hunt presence across 3 launches — all 3
   competitors launched 4+ times."

📊 Limits & pricing

1 credit per analyze_competitor call (same as the web UI)
Read-only tools (get_report*, list_my_reports) are unlimited
Reports persist forever by default (server-side storage) — use the job_id in any future session
Free tier: 3 analyses/month · Pro: 30/month · Team: unlimited (see pricing)

🛠️ Troubleshooting

AUTH_REQUIRED on every call: Token expired (~1h). Refresh from DevTools as described above.
INSUFFICIENT_CREDITS: Top up at /pricing or wait for monthly reset.
Tool responds but agent says "not connected": Fully restart the MCP client — Claude Desktop caches server state across reloads.
Still stuck?: Email iris@gingiris.com with your job_id and the failing call.

Built on the official Model Context Protocol · Source on GitHub.