structdoc-mcp 0.1.0

package/README.md

@@ -0,0 +1,33 @@
+# structdoc-mcp
+
+MCP server for **StructDoc** — turn any document (PDF/image) into structured data for AI agents. OCR, layout-to-Markdown, tables, key-value pairs, invoice/receipt fields. 100+ languages. Pay per call with x402 (USDC on Base/Solana) — no API keys.
+
+## Install
+
+```bash
+claude mcp add structdoc -e EVM_PRIVATE_KEY=0x<base-wallet-with-usdc> -- npx -y structdoc-mcp
+```
+
+Or configure any MCP client to run `npx -y structdoc-mcp` with env:
+
+- `EVM_PRIVATE_KEY` — Base wallet private key holding USDC, **and/or**
+- `SVM_PRIVATE_KEY` — Solana wallet private key (base58) holding USDC
+- `STRUCTDOC_BASE_URL` (optional) — default `https://structdoc-api.hp-vladic.workers.dev`
+
+## Tools
+
+| Tool | Does | Price |
+|---|---|---|
+| `document_to_markdown` | Layout → Markdown (headings, tables, key-value). RAG-ready. ≤20p | $0.25 |
+| `document_ocr` | OCR full-text, 100+ languages. ≤25p | $0.05 |
+| `extract_invoice` | Invoice fields (vendor, items, totals). ≤4p | $0.08 |
+| `extract_receipt` | Receipt fields (merchant, date, total). ≤2p | $0.04 |
+| `document_ocr_large` | OCR up to 100 pages | $0.20 |
+| `document_to_markdown_large` | Markdown up to 100 pages | $1.10 |
+| `structdoc_catalog` | Free machine-readable catalog | free |
+
+Each tool takes `{ url }` or `{ base64 }` (PDF/JPEG/PNG/BMP/TIFF/HEIF).
+
+Engine: Azure Document Intelligence v4.0. API: https://structdoc-api.hp-vladic.workers.dev
+
+MIT

package/index.mjs

@@ -0,0 +1,115 @@
+#!/usr/bin/env node
+// StructDoc MCP — turn any document into structured data for AI agents, paid per call via x402.
+//
+// Env (at least one):
+//   EVM_PRIVATE_KEY  Base wallet private key holding USDC (0x...)
+//   SVM_PRIVATE_KEY  Solana wallet private key holding USDC (base58)
+// Optional:
+//   EVM_RPC_URL      Base RPC, default https://mainnet.base.org
+//   STRUCTDOC_BASE_URL  API base, default https://structdoc-api.hp-vladic.workers.dev
+import { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
+import { StdioServerTransport } from '@modelcontextprotocol/sdk/server/stdio.js';
+import { z } from 'zod';
+import { x402Client, wrapFetchWithPayment } from '@x402/fetch';
+
+const BASE = (process.env.STRUCTDOC_BASE_URL ?? 'https://structdoc-api.hp-vladic.workers.dev').replace(/\/+$/, '');
+const RPC = process.env.EVM_RPC_URL ?? 'https://mainnet.base.org';
+
+let payingFetch = null;
+async function getFetch() {
+  if (payingFetch) return payingFetch;
+  const evmKey = process.env.EVM_PRIVATE_KEY;
+  const svmKey = process.env.SVM_PRIVATE_KEY;
+  if (!evmKey && !svmKey) {
+    throw new Error('Set EVM_PRIVATE_KEY (Base) and/or SVM_PRIVATE_KEY (Solana) — a wallet holding USDC to pay per call ($0.04-$1.10).');
+  }
+  const client = new x402Client();
+  if (evmKey) {
+    const { ExactEvmScheme } = await import('@x402/evm/exact/client');
+    const { privateKeyToAccount } = await import('viem/accounts');
+    client.register('eip155:*', new ExactEvmScheme(privateKeyToAccount(evmKey.startsWith('0x') ? evmKey : `0x${evmKey}`), { rpcUrl: RPC }));
+  }
+  if (svmKey) {
+    const { ExactSvmScheme } = await import('@x402/svm/exact/client');
+    const { createKeyPairSignerFromBytes } = await import('@solana/kit');
+    const { base58 } = await import('@scure/base');
+    client.register('solana:*', new ExactSvmScheme(await createKeyPairSignerFromBytes(base58.decode(svmKey))));
+  }
+  payingFetch = wrapFetchWithPayment(fetch, client);
+  return payingFetch;
+}
+
+async function post(path, payload) {
+  const f = await getFetch();
+  const res = await f(`${BASE}${path}`, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify(payload),
+  });
+  const text = await res.text();
+  if (!res.ok) throw new Error(`API ${res.status}: ${text.slice(0, 300)}`);
+  return text;
+}
+
+const asResult = text => ({ content: [{ type: 'text', text }] });
+const asError = e => ({ content: [{ type: 'text', text: `Error: ${String(e.message ?? e)}` }], isError: true });
+
+// url または base64 のどちらか必須の共通入力
+const docInput = {
+  url: z.string().optional().describe('URL of the document (PDF/JPEG/PNG/BMP/TIFF/HEIF)'),
+  base64: z.string().optional().describe('Base64-encoded document bytes (≤~6MB; use url for larger)'),
+};
+const pick = a => ({ url: a.url, base64: a.base64 });
+
+const server = new McpServer({ name: 'structdoc', version: '0.1.0' });
+
+server.tool(
+  'document_to_markdown',
+  'Convert any document (PDF/image) into agent-friendly MARKDOWN — headings, paragraphs, tables, key-value pairs. Best tool for feeding documents into an LLM or RAG pipeline. 100+ languages. Up to 20 pages (use document_to_markdown_large for more). Cost: $0.25 (x402/USDC).',
+  { ...docInput, features: z.string().optional().describe('Optional add-ons: comma-separated from languages,barcodes,keyValuePairs'), queryFields: z.string().optional().describe('Optional custom fields to extract, comma-separated (≤8)') },
+  async a => { try { return asResult(await post('/layout', { ...pick(a), features: a.features, queryFields: a.queryFields })); } catch (e) { return asError(e); } },
+);
+
+server.tool(
+  'document_ocr',
+  'OCR full-text extraction from a PDF or image in 100+ languages. Returns plain text content. Up to 25 pages. Cost: $0.05 (x402/USDC).',
+  { ...docInput, features: z.string().optional().describe('Optional: languages,barcodes') },
+  async a => { try { return asResult(await post('/read', { ...pick(a), features: a.features })); } catch (e) { return asError(e); } },
+);
+
+server.tool(
+  'extract_invoice',
+  'Extract structured invoice fields from a PDF/image invoice: vendor, customer, invoice id/date, line items, subtotal, tax, total, currency. Up to 4 pages. Cost: $0.08 (x402/USDC).',
+  { ...docInput, queryFields: z.string().optional().describe('Optional custom fields to extract, comma-separated (≤8)') },
+  async a => { try { return asResult(await post('/invoice', { ...pick(a), queryFields: a.queryFields })); } catch (e) { return asError(e); } },
+);
+
+server.tool(
+  'extract_receipt',
+  'Extract structured receipt fields from a receipt image/PDF: merchant, date/time, line items, subtotal, tax, tip, total, currency. Up to 2 pages. Cost: $0.04 (x402/USDC).',
+  { ...docInput },
+  async a => { try { return asResult(await post('/receipt', pick(a))); } catch (e) { return asError(e); } },
+);
+
+server.tool(
+  'document_ocr_large',
+  'OCR full-text extraction for large documents — up to 100 pages. Cost: $0.20 (x402/USDC).',
+  { ...docInput, features: z.string().optional() },
+  async a => { try { return asResult(await post('/read-large', { ...pick(a), features: a.features })); } catch (e) { return asError(e); } },
+);
+
+server.tool(
+  'document_to_markdown_large',
+  'Convert a large document (up to 100 pages) into Markdown for LLM/RAG use. Cost: $1.10 (x402/USDC).',
+  { ...docInput, features: z.string().optional() },
+  async a => { try { return asResult(await post('/layout-large', { ...pick(a), features: a.features })); } catch (e) { return asError(e); } },
+);
+
+server.tool(
+  'structdoc_catalog',
+  'Free: machine-readable catalog of StructDoc (endpoints, prices, supported formats). No payment required.',
+  {},
+  async () => { try { const r = await fetch(`${BASE}/catalog`); return asResult(await r.text()); } catch (e) { return asError(e); } },
+);
+
+await server.connect(new StdioServerTransport());

package/package.json

@@ -0,0 +1,21 @@
+{
+  "name": "structdoc-mcp",
+  "version": "0.1.0",
+  "description": "MCP server for StructDoc — turn any document (PDF/image) into structured data for AI agents: OCR, layout-to-Markdown, tables, invoice/receipt fields. Pays per call via x402 (USDC on Base/Solana) — no API keys.",
+  "type": "module",
+  "bin": { "structdoc-mcp": "index.mjs" },
+  "files": ["index.mjs", "README.md"],
+  "keywords": ["mcp", "x402", "document", "ocr", "pdf", "layout", "markdown", "rag", "invoice", "receipt", "modelcontextprotocol"],
+  "license": "MIT",
+  "homepage": "https://structdoc-api.hp-vladic.workers.dev",
+  "dependencies": {
+    "@modelcontextprotocol/sdk": "^1.12.0",
+    "@scure/base": "^2.2.0",
+    "@solana/kit": "^5.5.1",
+    "@x402/evm": "^2.18.0",
+    "@x402/fetch": "^2.18.0",
+    "@x402/svm": "^2.19.0",
+    "viem": "^2.21.26",
+    "zod": "^3.24.2"
+  }
+}