tavily-mcp

Instructions

Implementation Reference

• src/index.ts:496-511 (handler)

  The core handler function that executes the tavily-crawl tool logic by sending a POST request to Tavily's crawl API endpoint with the input parameters and the API key.

  async crawl(params: any): Promise<TavilyCrawlResponse> {
    try {
      const response = await this.axiosInstance.post(this.baseURLs.crawl, {
        ...params,
        api_key: API_KEY
      });
      return response.data;
    } catch (error: any) {
      if (error.response?.status === 401) {
        throw new Error('Invalid API key');
      } else if (error.response?.status === 429) {
        throw new Error('Usage limit exceeded');
      }
      throw error;
    }
  }

• src/index.ts:217-283 (schema)

  Input schema definition for the tavily-crawl tool, including all parameters, types, descriptions, defaults, and required fields, used in the ListTools response.

  {
    name: "tavily-crawl",
    description: "A powerful web crawler that initiates a structured web crawl starting from a specified base URL. The crawler expands from that point like a tree, following internal links across pages. You can control how deep and wide it goes, and guide it to focus on specific sections of the site.",
    inputSchema: {
      type: "object",
      properties: {
        url: {
          type: "string",
          description: "The root URL to begin the crawl"
        },
        max_depth: {
          type: "integer",
          description: "Max depth of the crawl. Defines how far from the base URL the crawler can explore.",
          default: 1,
          minimum: 1
        },
        max_breadth: {
          type: "integer",
          description: "Max number of links to follow per level of the tree (i.e., per page)",
          default: 20,
          minimum: 1
        },
        limit: {
          type: "integer",
          description: "Total number of links the crawler will process before stopping",
          default: 50,
          minimum: 1
        },
        instructions: {
          type: "string",
          description: "Natural language instructions for the crawler"
        },
        select_paths: {
          type: "array",
          items: { type: "string" },
          description: "Regex patterns to select only URLs with specific path patterns (e.g., /docs/.*, /api/v1.*)",
          default: []
        },
        select_domains: {
          type: "array",
          items: { type: "string" },
          description: "Regex patterns to select crawling to specific domains or subdomains (e.g., ^docs\\.example\\.com$)",
          default: []
        },
        allow_external: {
          type: "boolean",
          description: "Whether to allow following links that go to external domains",
          default: false
        },
        categories: {
          type: "array",
          items: { 
            type: "string",
            enum: ["Careers", "Blog", "Documentation", "About", "Pricing", "Community", "Developers", "Contact", "Media"]
          },
          description: "Filter URLs using predefined categories like documentation, blog, api, etc",
          default: []
        },
        extract_depth: {
          type: "string",
          enum: ["basic", "advanced"],
          description: "Advanced extraction retrieves more data, including tables and embedded content, with higher success but may increase latency",
          default: "basic"
        }
      },
      required: ["url"]
    }

• src/index.ts:381-399 (registration)

  Registration and dispatch logic in the CallToolRequestHandler switch statement that calls the crawl handler with parsed arguments and formats the response using formatCrawlResults.

  case "tavily-crawl":
    const crawlResponse = await this.crawl({
      url: args.url,
      max_depth: args.max_depth,
      max_breadth: args.max_breadth,
      limit: args.limit,
      instructions: args.instructions,
      select_paths: Array.isArray(args.select_paths) ? args.select_paths : [],
      select_domains: Array.isArray(args.select_domains) ? args.select_domains : [],
      allow_external: args.allow_external,
      categories: Array.isArray(args.categories) ? args.categories : [],
      extract_depth: args.extract_depth
    });
    return {
      content: [{
        type: "text",
        text: formatCrawlResults(crawlResponse)
      }]
    };

• src/index.ts:569-588 (helper)

  Helper function that formats the TavilyCrawlResponse into a human-readable string output for the tool response.

  function formatCrawlResults(response: TavilyCrawlResponse): string {
    const output: string[] = [];
    
    output.push(`Crawl Results:`);
    output.push(`Base URL: ${response.base_url}`);
    
    output.push('\nCrawled Pages:');
    response.results.forEach((page, index) => {
      output.push(`\n[${index + 1}] URL: ${page.url}`);
      if (page.raw_content) {
        // Truncate content if it's too long
        const contentPreview = page.raw_content.length > 200 
          ? page.raw_content.substring(0, 200) + "..." 
          : page.raw_content;
        output.push(`Content: ${contentPreview}`);
      }
    });
    
    return output.join('\n');
  }