Jina AI MCP Server

import { McpServer } from "@modelcontextprotocol/sdk/server/mcp.js"; import { z } from "zod"; import { JinaAIAPIClient } from "./client"; import * as fs from "fs/promises"; import * as path from "path"; export function registerAllTools(server: McpServer, jinaClient: JinaAIAPIClient) { // 1. Embeddings server.registerTool("embeddings", { title: "Jina AI Embeddings", description: "Creates an embedding vector representing the input text.", inputSchema: { input: z.union([z.string(), z.array(z.string())]).describe("The input text or texts to embed."), model: z.string().describe("The name of the model to use for embeddings, e.g., 'jina-embeddings-v2-base-en'."), }, }, async (input) => { const result = await jinaClient.embeddings(input); return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; }); // 2. Rerank server.registerTool("rerank", { title: "Jina AI Rerank", description: "Reranks a list of documents based on a query.", inputSchema: { query: z.string().describe("The query to use for reranking."), documents: z.array(z.string()).describe("A list of documents to rerank."), model: z.string().describe("The name of the model to use for reranking."), top_n: z.number().optional().describe("The number of documents to return."), }, }, async (input) => { const result = await jinaClient.rerank(input); return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; }); // 3. Read server.registerTool("read", { title: "Jina AI Reader", description: "Input a single website URL and get an LLM-friendly version of that single website.", inputSchema: { url: z.string().url().describe("The URL of the website to read."), options: z.record(z.any()).optional().describe("An object for additional headers like 'X-Target-Selector' or 'X-Return-Format'."), }, }, async ({ url, options }) => { const result = await jinaClient.read({ url }, options); return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; }); // 4. Search server.registerTool("search", { title: "Jina AI Search", description: "Given a search term, get an LLM-friendly version of all websites in the search results.", inputSchema: { query: z.string().describe("The search query."), options: z.record(z.any()).optional().describe("An object for additional search parameters and headers."), }, }, async ({ query, options }) => { const result = await jinaClient.search({ q: query }, options); return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; }); // 5. DeepSearch server.registerTool("deepsearch", { title: "Jina AI DeepSearch", description: "Combines web searching, reading, and reasoning for comprehensive investigation.", inputSchema: { messages: z.array(z.object({ role: z.enum(["user", "assistant"]), content: z.string(), })).describe("A list of messages forming the conversation so far."), model: z.string().describe("ID of the model to use, e.g., 'jina-deepsearch-v1'."), options: z.record(z.any()).optional().describe("Additional options like 'stream', 'reasoning_effort', etc."), }, }, async (input) => { const result = await jinaClient.deepsearch(input); return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; }); // 6. Segment server.registerTool("segment", { title: "Jina AI Segmenter", description: "Given a text, splits it into segments or counts tokens.", inputSchema: { content: z.string().describe("The text content to segment."), options: z.record(z.any()).optional().describe("An object for options like 'tokenizer', 'return_chunks', etc."), }, }, async (input) => { const result = await jinaClient.segment(input); return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; }); // 7. Classify server.registerTool("classify", { title: "Jina AI Classifier", description: "Zero-shot classification for text.", inputSchema: { input: z.array(z.string()).describe("Array of text inputs for classification."), labels: z.array(z.string()).describe("List of labels for classification."), model: z.string().describe("Model to use, e.g., 'jina-embeddings-v3' for text."), }, }, async (input) => { const result = await jinaClient.classify(input); return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; }); // 8. Get Help server.registerTool("get_help", { title: "Get Jina AI API Help", description: "Returns the full content of the jina-docs.md documentation.", inputSchema: {}, }, async () => { try { const docPath = path.resolve(__dirname, '../../jina-docs.md'); const content = await fs.readFile(docPath, 'utf-8'); return { content: [{ type: "text", text: content }] }; } catch (error) { console.error("Error reading help file:", error); return { content: [{ type: "text", text: "Error: Could not load the help documentation." }] }; } }); }