@brave/brave-search-mcp-server/dist/tools/llm_context/index.js

Brave Search MCP Server: web results, images, videos, rich results, AI summaries, and more.

import { z } from 'zod';
import API from '../../BraveAPI/index.js';
import { RequestParamsSchema, RequestHeadersSchema, LlmContextInputSchema, } from './schemas/input.js';
import { LlmContextSearchApiResponseSchema } from './schemas/output.js';
export const name = 'brave_llm_context';
export const annotations = {
    title: 'Brave LLM Context',
    openWorldHint: true,
};
export const description = `
    Retrieves pre-extracted, relevance-ranked web content using Brave's LLM Context API, optimized for AI agents, LLM grounding, and RAG pipelines. Unlike a traditional web search that returns links and short descriptions, this tool returns the actual substance of matching pages — text chunks, tables, code blocks, and structured data — so the model can reason over it directly.

    When to use:
        - Grounding answers in fresh, relevant web content (RAG)
        - Giving an AI agent ready-to-use page content from a single search call
        - Question answering and fact-checking against current sources
        - Gathering source material for research without manually fetching pages
        - When you need the contents of pages, not just titles, descriptions, and URLs

    When relaying results in markdown-supporting environments, cite the source URLs from the "sources" map.
`.trim();
export const execute = async (params) => {
    const parsedParams = RequestParamsSchema.parse(params);
    const parsedHeaders = RequestHeadersSchema.parse(params);
    const response = await API.issueRequest('llmContext', parsedParams, parsedHeaders);
    const { success, data, error } = LlmContextSearchApiResponseSchema.safeParse(response);
    const payload = success ? data : z.treeifyError(error);
    return {
        content: [{ type: 'text', text: JSON.stringify(payload) }],
        isError: !success,
        structuredContent: payload,
    };
};
export const register = (mcpServer) => {
    mcpServer.registerTool(name, {
        title: name,
        description: description,
        inputSchema: LlmContextInputSchema.shape,
        outputSchema: LlmContextSearchApiResponseSchema.shape,
        annotations: annotations,
    }, execute);
};
export default {
    name,
    description,
    annotations,
    inputSchema: LlmContextInputSchema.shape,
    outputSchema: LlmContextSearchApiResponseSchema.shape,
    execute,
    register,
};