docuprox-mcp
Enables AI clients like Claude to process documents (invoices, passports, etc.) via the DocuProx API, with tools for submitting jobs, checking status, and retrieving results.
README
docuprox-mcp
MCP (Model Context Protocol) server for the DocuProx document-processing API. Exposes DocuProx as tools that any MCP-compatible AI client can call — Claude Desktop, Claude Code, Cursor, Windsurf, and more.
Quickstart (npx — no install required)
DOCUPROX_API_KEY=your_key DOCUPROX_BASE_URL=https://api.docuprox.com/v1 npx docuprox-mcp
Requires Node.js ≥ 18.
Configuration
| Variable | Required | Description |
|---|---|---|
DOCUPROX_API_KEY |
yes | Your DocuProx API key |
DOCUPROX_BASE_URL |
yes | Base URL of the DocuProx API (e.g. https://api.docuprox.com/v1) |
Connect to an MCP Client
Claude Desktop
macOS — ~/Library/Application Support/Claude/claude_desktop_config.json
Windows — %APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"docuprox": {
"command": "npx",
"args": ["docuprox-mcp"],
"env": {
"DOCUPROX_API_KEY": "your_api_key_here",
"DOCUPROX_BASE_URL": "https://api.docuprox.com/v1"
}
}
}
}
Claude Code (CLI)
claude mcp add docuprox -e DOCUPROX_API_KEY=your_api_key_here -e DOCUPROX_BASE_URL=https://api.docuprox.com/v1 -- npx docuprox-mcp
Cursor / Windsurf
Add to your MCP settings (.cursor/mcp.json or equivalent):
{
"mcpServers": {
"docuprox": {
"command": "npx",
"args": ["docuprox-mcp"],
"env": {
"DOCUPROX_API_KEY": "your_api_key_here",
"DOCUPROX_BASE_URL": "https://api.docuprox.com/v1"
}
}
}
}
Available Tools
process_job
Submit a document for asynchronous processing. Returns a job_id to track with poll_job or job_results.
| Argument | Type | Required | Description |
|---|---|---|---|
file_path |
string | yes | Path to file on disk (jpg, png, pdf, zip) |
template_id |
string | no | UUID of a DocuProx template |
document_type |
string | no* | Document category — required when no template_id |
prompt_json |
object | no | Custom extraction schema |
custom_instructions |
string | no | Free-text guidance for the AI |
static_values |
object | no | Key-value overrides for STATIC template placeholders |
Example: "Submit the invoice at
/tmp/invoice.pdfusing template123e4567-e89b-12d3-a456-426614174000"
process_agent
Submit a document for synchronous extraction. Blocks until done and returns results directly — no polling needed. Credits are automatically refunded on failure.
| Argument | Type | Required | Description |
|---|---|---|---|
file_path |
string | yes | Path to file on disk |
prompt_json |
object | yes | Extraction schema |
document_type |
string | yes | Document category |
custom_instructions |
string | no | Free-text guidance |
static_values |
object | no | Key-value STATIC overrides |
Example: "Extract passport data from
/home/user/passport.jpg. Use document_type='passport' and prompt_json{ 'name': 'full name', 'dob': 'date of birth' }"
job_status
Check the status of an async job.
| Argument | Type | Required | Description |
|---|---|---|---|
job_id |
string | yes | UUID from process_job |
Returns: { job_id, status } where status is one of:
NEW → UNZIP FILE → UNZIP FILE SUCCESS / UNZIP FILE FAILED → PROCESS IMAGE → PROCESS IMAGE SUCCESS / PROCESS IMAGE FAILED → SUCCESS / FAILED
poll_job
Block until a job finishes (or times out). Retries job_status internally on a timer.
| Argument | Type | Required | Default | Description |
|---|---|---|---|---|
job_id |
string | yes | — | UUID from process_job |
interval_ms |
number | no | 3000 |
Poll interval in ms |
timeout_ms |
number | no | 300000 |
Max wait in ms (5 min) |
Example: "Submit
/tmp/batch.zipthen wait for the result" — Claude will callprocess_jobthenpoll_jobautomatically.
job_results
Fetch the extracted results of a completed async job.
| Argument | Type | Required | Default | Description |
|---|---|---|---|---|
job_id |
string | yes | — | UUID from process_job |
result_format |
string | no | "json" |
Output format: "json" or "csv" |
Example: "Get the results for job
0869d5ec-5cfe-4960-878d-8b4ec1900726in CSV format"
How File Uploads Work
Every tool that accepts a file_path:
- Resolves it to an absolute path
- Reads the file into a
Buffer - Detects MIME type from the file extension
- Sends it as
multipart/form-datawith field nameactual_image
The AI never sees or handles base64 — just pass a local file path.
Error Handling
| Error | Cause |
|---|---|
| File not found | file_path does not exist — a clear message with the resolved path is returned |
| API error | HTTP status + response body surfaced in the tool result |
| Invalid arguments | Zod validation fires before any API call |
| Credit failure | 402/403 from /process-agent surfaced clearly |
Local Development
# Clone and install
git clone <repo>
cd docuprox-mcp
npm install
# Run in dev mode (no build step)
DOCUPROX_API_KEY=test DOCUPROX_BASE_URL=http://localhost:5000/v1 npm run dev
# Build
npm run build
# Inspect with MCP Inspector
npx @modelcontextprotocol/inspector node dist/index.js
Project Structure
docuprox-mcp/
├── index.ts ← MCP server entry point (stdio transport)
├── config.ts ← Env-var config loader
├── client.ts ← Axios API client (multipart/form-data uploads)
├── tools.ts ← Zod schemas + MCP tool definitions
├── handlers.ts ← Maps tool names → client calls
├── package.json
└── tsconfig.json
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。