Jina AI Remote MCP Server

Instructions

Implementation Reference

• src/tools/jina-tools.ts:1120-1228 (handler)

  The core handler function for the 'extract_pdf' tool. It validates inputs, makes a POST request to 'https://svip.jina.ai/extract-pdf' with the provided arXiv ID or PDF URL, processes the JSON response containing extracted figures/tables/equations as base64 images with metadata, limits to first 20 items, constructs content items with YAML metadata and PNG images, and handles errors.
  async ({ id, url, max_edge, type }: { id?: string; url?: string; max_edge: number; type?: string }) => { try { const props = getProps(); const tokenError = checkBearerToken(props.bearerToken); if (tokenError) { return tokenError; } if (!id && !url) { return createErrorResponse("Either 'id' (arXiv paper ID) or 'url' (PDF URL) is required"); } // Build request body const requestBody: Record<string, any> = {}; if (id) requestBody.id = id; if (url) requestBody.url = url; if (max_edge) requestBody.max_edge = max_edge; if (type) requestBody.type = type; const response = await fetch('https://svip.jina.ai/extract-pdf', { method: 'POST', headers: { 'Accept': 'application/json', 'Content-Type': 'application/json', 'Authorization': `Bearer ${props.bearerToken}`, }, body: JSON.stringify(requestBody), }); if (!response.ok) { return handleApiError(response, "PDF extraction"); } const data = await response.json() as { id: string; floats: Array<{ type: string; number: string; caption: string; page: number; image: string; width: number; height: number; }>; meta: { latency: number; num_floats: number; num_pages: number; total_bytes: number; credits: number; tokens: number; }; }; // Limit floats to prevent large responses const maxFloats = 20; const totalFloats = data.floats.length; const floatsToReturn = data.floats.slice(0, maxFloats); // Return each float as an image with metadata const contentItems: Array<{ type: 'text'; text: string } | { type: 'image'; data: string; mimeType: string }> = []; // Add summary metadata const summaryMeta: Record<string, any> = { id: data.id, num_floats: data.meta.num_floats, num_pages: data.meta.num_pages, latency_ms: data.meta.latency }; if (totalFloats > maxFloats) { summaryMeta.returned_floats = maxFloats; summaryMeta.truncated = true; summaryMeta.note = `Showing first ${maxFloats} of ${totalFloats} floats. Use 'type' parameter to filter by specific types.`; } contentItems.push({ type: "text" as const, text: yamlStringify(summaryMeta), }); // Add each float as an image with its metadata for (const float of floatsToReturn) { // Add metadata for this float contentItems.push({ type: "text" as const, text: yamlStringify({ type: float.type, number: float.number, caption: float.caption, page: float.page, dimensions: `${float.width}x${float.height}` }), }); // Add the image contentItems.push({ type: "image" as const, data: float.image, mimeType: "image/png", }); } return { content: contentItems, }; } catch (error) { return createErrorResponse(`Error: ${error instanceof Error ? error.message : String(error)}`); } },
• src/tools/jina-tools.ts:1115-1119 (schema)

  Zod schema defining the input parameters for the 'extract_pdf' tool: optional arXiv ID or PDF URL (one required), max_edge for image size, and type filter for figures/tables/equations.
  id: z.string().optional().describe("arXiv paper ID (e.g., '2301.12345' or 'hep-th/9901001'). Either id or url is required."), url: z.string().url().optional().describe("Direct PDF URL. Either id or url is required."), max_edge: z.number().default(1024).describe("Maximum edge size for extracted images in pixels (default: 1024)"), type: z.string().optional().describe("Filter by float types (comma-separated): figure, table, equation. If not specified, returns all types.") },
• src/tools/jina-tools.ts:1110-1230 (registration)

  The conditional registration of the 'extract_pdf' tool using server.tool within the registerJinaTools function, only if enabled.
  if (isToolEnabled("extract_pdf")) { server.tool( "extract_pdf", "Extract figures, tables, and equations from PDF documents using layout detection. Perfect for extracting visual elements from academic papers on arXiv or any PDF URL. Returns base64-encoded images of detected elements with metadata.", { id: z.string().optional().describe("arXiv paper ID (e.g., '2301.12345' or 'hep-th/9901001'). Either id or url is required."), url: z.string().url().optional().describe("Direct PDF URL. Either id or url is required."), max_edge: z.number().default(1024).describe("Maximum edge size for extracted images in pixels (default: 1024)"), type: z.string().optional().describe("Filter by float types (comma-separated): figure, table, equation. If not specified, returns all types.") }, async ({ id, url, max_edge, type }: { id?: string; url?: string; max_edge: number; type?: string }) => { try { const props = getProps(); const tokenError = checkBearerToken(props.bearerToken); if (tokenError) { return tokenError; } if (!id && !url) { return createErrorResponse("Either 'id' (arXiv paper ID) or 'url' (PDF URL) is required"); } // Build request body const requestBody: Record<string, any> = {}; if (id) requestBody.id = id; if (url) requestBody.url = url; if (max_edge) requestBody.max_edge = max_edge; if (type) requestBody.type = type; const response = await fetch('https://svip.jina.ai/extract-pdf', { method: 'POST', headers: { 'Accept': 'application/json', 'Content-Type': 'application/json', 'Authorization': `Bearer ${props.bearerToken}`, }, body: JSON.stringify(requestBody), }); if (!response.ok) { return handleApiError(response, "PDF extraction"); } const data = await response.json() as { id: string; floats: Array<{ type: string; number: string; caption: string; page: number; image: string; width: number; height: number; }>; meta: { latency: number; num_floats: number; num_pages: number; total_bytes: number; credits: number; tokens: number; }; }; // Limit floats to prevent large responses const maxFloats = 20; const totalFloats = data.floats.length; const floatsToReturn = data.floats.slice(0, maxFloats); // Return each float as an image with metadata const contentItems: Array<{ type: 'text'; text: string } | { type: 'image'; data: string; mimeType: string }> = []; // Add summary metadata const summaryMeta: Record<string, any> = { id: data.id, num_floats: data.meta.num_floats, num_pages: data.meta.num_pages, latency_ms: data.meta.latency }; if (totalFloats > maxFloats) { summaryMeta.returned_floats = maxFloats; summaryMeta.truncated = true; summaryMeta.note = `Showing first ${maxFloats} of ${totalFloats} floats. Use 'type' parameter to filter by specific types.`; } contentItems.push({ type: "text" as const, text: yamlStringify(summaryMeta), }); // Add each float as an image with its metadata for (const float of floatsToReturn) { // Add metadata for this float contentItems.push({ type: "text" as const, text: yamlStringify({ type: float.type, number: float.number, caption: float.caption, page: float.page, dimensions: `${float.width}x${float.height}` }), }); // Add the image contentItems.push({ type: "image" as const, data: float.image, mimeType: "image/png", }); } return { content: contentItems, }; } catch (error) { return createErrorResponse(`Error: ${error instanceof Error ? error.message : String(error)}`); } }, ); }
• src/index.ts:100-100 (registration)

  Call to registerJinaTools which includes the 'extract_pdf' tool registration, passing the MCP server instance and props.
  registerJinaTools(server, () => currentProps, enabledTools);
• src/index.ts:24-24 (registration)

  'extract_pdf' listed in ALL_TOOLS array used for tool filtering and enabling.
  "sort_by_relevance", "deduplicate_strings", "deduplicate_images", "extract_pdf"