CompetLab MCP Server is CompetLab platform's competitive intelligence interface for AI agents, built on the Model Context Protocol. It enables AI agents to access CompetLab's competitive intelligence data, including dashboards, historical data, alerts, strategic briefs, and the project strategy ticket board.
Primary Use As more and more B2B buyers ask AI before Google, CompetLab monitors competitors' performance across 6 dimensions, with key dimensions including: AI Visibility, i.e., which brands ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews recommend when buyers ask questions, and whether you are among the core recommendations; AI Sources, i.e., the pages Perplexity and Google AI Overviews retrieve when generating these answers, and whether you are mentioned. It also covers price intelligence, content intelligence, technology and trust profiles, and more.
Primary Capabilities This MCP server provides 48 tools, of which 40 are read-only, 3 are asynchronous scan starters (start_tech_stack_scan, start_trust_signals_scan, start_agent_adoption_scan), which create scan records; 5 write to the project strategy ticket board (create_ticket, update_ticket, move_ticket, delete_ticket, add_ticket_comment), requiring a read_write API key. The tools cover projects and competitors, price intelligence, content intelligence, technology and trust profiles, strategic briefs, strategy tickets, alerts and schedules, and free tools that require no project setup. Free tools can run against any public domain and do not require a projectId; synchronous tools return immediately, while asynchronous scans return a scanId and require polling every 5–10 seconds. Paginated tools support the page and limit parameters, and pagination.hasMore indicates whether there are more pages. Dashboards related to AI Visibility and AI Sources, as well as the technology and trust dashboards, are returned in a compact view by default, with market maps, page lists, and brand lists presented page by page, along with a Page object describing the total number of rows; passing view=full returns all rows at once. The response begins with a readingGuide, explaining the reading rules for each field. The response directly passes through the raw results of the CompetLab API, and server instructions tell the agent how to interpret them, where null means CompetLab did not measure that value, rather than zero or "no".
How to Use Supports any MCP-compatible client, including Claude Desktop, Claude Web, VS Code (Copilot), Cursor, Windsurf, Cline, etc. There are two connection methods:
- Remote server (recommended): the server URL is https://mcp.competlab.com/mcp，通过 API key authentication via the CL-API-Key request header or the api_key query parameter. Configuration file paths and field names differ across clients; for example, Cursor uses .cursor/mcp.json, VS Code uses .vscode/mcp.json (using servers instead of mcpServers, and supporting ${input:id} secure input prompts), Windsurf uses ~/.codeium/windsurf/mcp_config.json (using serverUrl instead of url), and Cline uses cline_mcp_settings.json. Claude Desktop and Claude Web only support URL-based authentication (no custom request headers), so the api_key query parameter must be used.
- Local server (stdio): runs via stdin/stdout, suitable for Claude Desktop, Glama, or environments that prefer the stdio transport. It must be added to claude_desktop_config.json in the Claude Desktop configuration. The server reads JSON-RPC from stdin and writes responses to stdout. The repository's examples/ directory provides ready-to-use configuration examples for each client.
Getting an API key: register at app.competlab.com/register (free 14-day trial, no credit card required), go to Organization Settings > API Keys to create a new key; keys start with cl_live_. One API key covers the entire organization.
Use cases It lets AI agents answer questions such as: which companies AI models recommend in my category, and whether I am among them; which pages Perplexity and Google AI Overviews read for buyer questions that mention competitors rather than me; what changed on competitors' pricing pages this week; what a strategy brief should fix first; which tickets the latest brief opened and their board status; how the AI market map has changed over the past 3 months; compare the content strategies of all tracked competitors; which key alerts were triggered in the past 7 days; which competitors have better security headers than we do; run a techstack scan on stripe.com; according to robots.txt, which AI assistants can access openai.com; use cleanHtml to crawl and summarize a g2.com list page, etc.
Limits The fetch_url tool is rate-limited by API key to 60 requests/minute, stricter than the default 1000/minute of other free tools. The three start_*_scan tools create scan records under the account (they do not edit existing data); the five strategy ticket write tools require a read_write key, and read-only keys will be rejected. A read-only key can read all organization data and can operate the strategy ticket board on any plan and after expiration; however, calls to fetch or scan sites on the CompetLab side require a valid plan, otherwise they return 402 subscription_required. MCP access is provided with all CompetLab plans and the free 14-day trial. The documentation and configuration in the repository are licensed under MIT; the CompetLab MCP server and platform are commercial software. Other details such as charges and pricing were not provided.
