竞争情报MCP服务器

CompetLab MCP Server

Competitive intelligence for AI agents — see where AI sends your buyers, and what to do about it.

More B2B buyers are asking AI before they Google. CompetLab monitors competitors across 6 dimensions — including AI Visibility, which tracks which brands ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews recommend, and AI Sources, the pages Perplexity and Google AI Overviews read when they answer your buyers' questions. This MCP server gives your AI agent access to all of it: dashboards, historical data, alerts, the Strategic Briefing, and the project's Strategic Tickets board.

Supported Clients

Works with any MCP-compatible client:

• Claude Desktop / Claude Web
• Claude Code
• Cursor
• VS Code (Copilot)
• Windsurf
• Cline

Quick Start

Two ways to connect — pick the one that fits your setup:

Get your API key: app.competlab.com > Organization Settings > API Keys

Option 1: Remote Server (recommended)

Server URL: https://mcp.competlab.com/mcp Auth: API key via CL-API-Key header (or api_key query parameter)

Claude Code

claude mcp add --transport http \
  --header "CL-API-Key: YOUR_COMPETLAB_API_KEY" \
  competlab https://mcp.competlab.com/mcp

Cursor

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "competlab": {
      "url": "https://mcp.competlab.com/mcp",
      "headers": {
        "CL-API-Key": "YOUR_COMPETLAB_API_KEY"
      }
    }
  }
}

VS Code

Add to .vscode/mcp.json:

{
  "inputs": [
    {
      "type": "promptString",
      "id": "competlab-api-key",
      "description": "CompetLab API Key (starts with cl_live_)",
      "password": true
    }
  ],
  "servers": {
    "competlab": {
      "type": "http",
      "url": "https://mcp.competlab.com/mcp",
      "headers": {
        "CL-API-Key": "${input:competlab-api-key}"
      }
    }
  }
}

Note: VS Code uses "servers" (not "mcpServers") and supports secure input prompts via ${input:id}.

Windsurf

Add to ~/.codeium/windsurf/mcp_config.json:

{
  "mcpServers": {
    "competlab": {
      "serverUrl": "https://mcp.competlab.com/mcp",
      "headers": {
        "CL-API-Key": "YOUR_COMPETLAB_API_KEY"
      }
    }
  }
}

Note: Windsurf uses "serverUrl" (not "url").

Cline

Add to cline_mcp_settings.json (or configure via Cline UI > Installed > Advanced MCP Settings):

{
  "mcpServers": {
    "competlab": {
      "url": "https://mcp.competlab.com/mcp",
      "headers": {
        "CL-API-Key": "YOUR_COMPETLAB_API_KEY"
      },
      "disabled": false
    }
  }
}

Claude Desktop / Claude Web

Claude Desktop and Claude Web only support URL-based auth (no custom headers). Use the api_key query parameter:

Go to Settings > MCP and add the server with this URL:

https://mcp.competlab.com/mcp?api_key=YOUR_COMPETLAB_API_KEY

Option 2: Local Server (stdio)

Run the server locally via stdin/stdout. Useful for Claude Desktop, Glama, or environments that prefer stdio transport.

git clone https://github.com/competlab/competlab-mcp-server.git
cd competlab-mcp-server
npm install
npm run build

Claude Code

claude mcp add --transport stdio \
  --env COMPETLAB_API_KEY=YOUR_COMPETLAB_API_KEY \
  competlab node dist/index.js

Claude Desktop

Add to your Claude Desktop config (claude_desktop_config.json):

{
  "mcpServers": {
    "competlab": {
      "command": "node",
      "args": ["dist/index.js"],
      "cwd": "/path/to/competlab-mcp-server",
      "env": {
        "COMPETLAB_API_KEY": "YOUR_COMPETLAB_API_KEY"
      }
    }
  }
}

Generic stdio

COMPETLAB_API_KEY=YOUR_COMPETLAB_API_KEY node dist/index.js

The server reads JSON-RPC from stdin and writes responses to stdout.

See examples/ for ready-to-paste config files for each client.

What is CompetLab?

Competitive intelligence for the AI era: 14 dimensions — 6 monitored continuously, plus 8 leading-edge dimensions researched for a Strategic Briefing every month, or every two weeks on Process. The six monitored dimensions:

AI Visibility answers who AI recommends — which brands ChatGPT, Claude, Gemini, Perplexity and Google AI Overviews name and recommend when your buyers ask, and whether you are in the core. AI Sources is its companion: the pages Perplexity and Google AI Overviews retrieve on the way to those answers, and whether they name you.

Start free trial (14 days, no credit card) | Learn more

Available Tools

48 tools. 40 are read-only; 3 are async-scan starters that create a scan record (start_tech_stack_scan, start_trust_signals_scan, start_agent_adoption_scan); 5 write to the project's Strategic Tickets board (create_ticket, update_ticket, move_ticket, delete_ticket, add_ticket_comment) and need a read_write API key.

Projects & Competitors

AI Visibility

AI Sources

Positioning

Pricing Intelligence

Content Intelligence

Tech & Trust Profile

Strategic Briefing

Strategic Tickets

The project's board — the work the team has decided to do, with an owner, a column and a thread. The same tickets the team sees in the app, in five fixed columns: triage, todo, in_progress, done, dismissed. Every move in a Strategic Briefing lands here — as a new ticket in triage, most important first, or on the ticket already there for that work — and a later edition comments on tickets already there when it measured something about them. Every tool that takes a ticket ID also takes the ticket's number as a person writes it, #14. A read key lists and reads tickets; the tools that write need a read_write key. A key reads everything its organization has and works its Strategic Tickets board on any plan and after one ends; the calls that fetch or scan a site on CompetLab's side need a plan (402 subscription_required otherwise).

Alerts & Schedules

Free Tools (no project setup required)

Run these against any public domain — no projectId needed. The sync tools return immediately; the async scans return a scanId you poll every 5–10 seconds.

All paginated tools accept page and limit parameters. Check pagination.hasMore in the response to fetch more pages.

The AI Visibility and AI Sources dashboards and check details, the AI Visibility history, and the Tech & Trust dashboard answer in a compact view by default: the market map, the pages list and the brands list come one page at a time, with your own row — and, on the market map, every tracked competitor's — always on the page and a *Page object (offset, limit, total, hasMore) saying how many rows there are. Pass view=full for every row in one response. Every one of these responses opens with readingGuide, the reading rules for its fields.

Responses pass through from the CompetLab API unchanged, and the server's instructions tell your agent how to read them — above all, null means CompetLab did not measure a value, never zero or "no".

Example Prompts

Once connected, try asking your AI agent:

• "Which companies do the AI models recommend in my category — and am I one of them?"
• "Which pages do Perplexity and Google AI Overviews read for my buyers' questions that name my competitors but not me?"
• "What changed on my competitors' pricing pages this week?"
• "Show me the strategic briefing — what should I fix first?"
• "Which tickets did the latest briefing open, and where do they stand on our board?"
• "How has the AI market map moved over the last 3 months?"
• "Compare content strategies across all my tracked competitors"
• "What critical alerts fired in the last 7 days?"
• "Which competitors have better security headers than us?"
• "Run a tech-stack scan on stripe.com — what are they using?"
• "Which AI assistants can reach openai.com, according to its robots.txt?"
• "Fetch g2.com/some-listing with cleanHtml and summarize the page"

See examples/prompts.md for more prompts organized by use case.

Authentication

Getting an API key

1. Sign up at app.competlab.com/register (free 14-day trial, no credit card)
2. Go to Organization Settings > API Keys
3. Create a new key — it starts with cl_live_

Two authentication methods

One API key covers your entire organization. Most tools are read-only; the three start_*_scan tools create scan records under your account (no edits to existing data), and the five Strategic Tickets write tools change the project's board — they need a read_write key, and a read key is refused on them. The fetch_url tool is rate-limited at 60 req/min per API key (tighter than the 1000/min default for other free tools).

Pricing

MCP access comes with every CompetLab plan and with the free 14-day trial.

Troubleshooting

Links

• MCP Server Documentation
• REST API Reference
• TypeScript SDK (npm install @competlab/sdk)
• Privacy Policy
• Start Free Trial

Support

• Bug reports: GitHub Issues
• Email: support@competlab.com
• Documentation: competlab.com/developers

License

MIT (covers documentation and configs in this repo) — see LICENSE

The CompetLab MCP server and platform are commercial software. See competlab.com/terms-and-conditions.

Built by the CompetLab team. Competitive intelligence for the AI era.