x402-mcp
MCP server for the x402 protocol that lets AI agents discover and call payment-gated HTTP APIs automatically.
README
x402-mcp
MCP server for the x402 protocol. Lets AI agents (Claude, Cursor, etc.) discover and call payment-gated HTTP APIs automatically.
Repository: merkleworks/x402-mcp
MCP Server
This repository provides a Model Context Protocol (MCP) server that enables AI agents to call payment-gated HTTP APIs using the x402 protocol.
Tools exposed:
x402_http_request— Execute HTTP requests to APIs requiring payment via the x402 protocol. If the server returns 402, the tool parses the challenge, constructs the payment, retries with proof, and returns the final response.
Example agent flow:
agent → tool call → x402_http_request → 402 challenge → payment → retry → response
Automatic discovery: The server is discoverable by MCP-aware frameworks without manual registry submission:
- Repository descriptor: mcp.json at the repo root (name, description,
install: npx x402-mcp, tools). - npm: Run with
npx x402-mcp; package keywords includemcp,model-context-protocol,ai-tools,x402. - Well-known: .well-known/mcp.json can be hosted at
https://demo.x402.merkleworks.io/.well-known/mcp.jsonfor domain-based discovery.
Overview of x402
x402 is a protocol that gates HTTP APIs behind micropayments using HTTP 402 Payment Required:
- Request — Client calls the API.
- 402 challenge — Server responds with
402 Payment Requiredand anX402-Challengeheader (nonce UTXO, price, payee, expiry). - Payment — Client builds a BSV transaction, delegates fees (optional), and broadcasts.
- Retry — Client retries the request with an
X402-Proofheader. - Response — Server verifies the proof and returns the API response.
This flow lets APIs charge per request in satoshis without pre-registration or cards.
What this MCP server does
This server exposes Model Context Protocol (MCP) tools so AI agents can:
paid_http_request— Send an HTTP request; if the server returns 402, run the full x402 flow (parse challenge → pay → retry with proof) and return the final response. Primary tool for calling payment-gated APIs.discover_x402_api— List payable endpoints (from/.well-known/x402or by probing for 402).pay_x402_endpoint— Call a known x402 endpoint with full control (method, body, headers, delegator).parse_x402_challenge— Decode anX402-Challengeheader.verify_x402_proof— Check anX402-Proofagainst a challenge and request.
Agents use paid_http_request for a single “call this URL; pay if 402” step.
Installation
Run with a single command (recommended):
npx x402-mcp
Or install globally:
npm install -g x402-mcp
x402-mcp
From source:
git clone https://github.com/merkleworks/x402-mcp.git
cd x402-mcp
npm install
npm run build
npm start
Usage example
Agent → tool call → x402_http_request → x402 flow (if 402) → API response
- The agent calls the tool with
url,method, and optionalheadersandbody. - The server sends the HTTP request.
- If the response is 402, the server parses the challenge, builds the payment, delegates/broadcasts, then retries with the proof.
- The tool returns the final response (status, body, and payment txid if paid).
Example tool call
Tool: x402_http_request
Input schema:
| Field | Type | Required | Description |
|---|---|---|---|
url |
string | yes | Full URL of the API endpoint |
method |
string | no | HTTP method (default: GET) |
headers |
object | no | Optional HTTP headers (key-value strings) |
body |
string or object | no | Optional request body (string or JSON object) |
Example (GET):
{
"tool": "x402_http_request",
"input": {
"url": "https://demo.x402.merkleworks.io/v1/expensive",
"method": "GET"
}
}
Example (POST with JSON body):
{
"tool": "x402_http_request",
"input": {
"url": "https://api.example.com/v1/query",
"method": "POST",
"headers": { "Content-Type": "application/json" },
"body": { "query": "example" }
}
}
More examples: examples/, including example_agent_call.json.
Connecting from AI clients
Cursor
Add to .cursor/mcp.json:
{
"mcpServers": {
"x402": {
"command": "npx",
"args": ["@merkleworks/x402-mcp"]
}
}
}
Or with a local build:
{
"mcpServers": {
"x402": {
"command": "node",
"args": ["/path/to/x402-mcp/dist/server.js"]
}
}
}
Claude Desktop / Claude Code
Add to your MCP config (e.g. claude_desktop_config.json):
{
"mcpServers": {
"x402": {
"command": "npx",
"args": ["@merkleworks/x402-mcp"]
}
}
}
Generic MCP client (Node.js)
import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";
const transport = new StdioClientTransport({
command: "npx",
args: ["@merkleworks/x402-mcp"],
});
const client = new Client({ name: "my-agent", version: "1.0.0" });
await client.connect(transport);
const result = await client.callTool("paid_http_request", {
url: "https://demo.x402.merkleworks.io/v1/expensive",
method: "GET",
});
console.log(result);
Tools reference
| Tool | Description |
|---|---|
paid_http_request |
Execute HTTP request; if 402, run x402 flow and return final response |
discover_x402_api |
Discover payable endpoints (/.well-known/x402 or probe) |
pay_x402_endpoint |
Call x402 endpoint with full control (delegator, broadcast URL, etc.) |
parse_x402_challenge |
Decode X402-Challenge header |
verify_x402_proof |
Verify X402-Proof against challenge and request |
Protocol summary
- Client requests a protected endpoint.
- Server responds 402 with
X402-Challenge. - Client builds a BSV tx, optionally uses a delegator for fees, then broadcasts.
- Client retries with
X402-Proof. - Server verifies and returns the response.
This MCP server implements that flow so agents can pay for API calls without manual steps.
Configuration
pay_x402_endpoint(and thuspaid_http_requestwhen it pays) use the same origin as the request URL for the delegator by default. Override withdelegator_url/broadcast_urlwhen usingpay_x402_endpointdirectly.- Discovery manifest format: protocol spec and
/.well-known/x402.
Examples
- example_agent_call.json — Example
paid_http_requestcall for agents. - discover-and-pay — Discover endpoints then pay and call the first. Run:
npx tsx examples/discover-and-pay.ts https://your-x402-api.example.com
MCP registry (mcp.json)
The repo includes mcp.json for MCP registries (e.g. mcp.so): name, description, repository, install command, and tool list.
License
Apache-2.0
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。