HTTPayer MCP
Enables AI agents to call x402-gated APIs using a central credit balance, abstracting away blockchain complexity and payment proofs. It provides tools to fetch data from payment-required endpoints, check usage balances, and simulate transaction costs.
README
@httpayer/mcp
MCP (Model Context Protocol) server for HTTPayer. Lets AI agents call x402-enabled APIs using credit balance — no wallets, no blockchain, no Web3 knowledge required.
- Dashboard & API keys: app.httpayer.com
- npm: @httpayer/mcp
- GitHub: httpayer/mcp
Quickstart
With an AI agent (recommended)
Paste this into any MCP-compatible agent (Claude Code, Cursor, Windsurf, OpenCode...):
Set up https://httpayer.com/skill.md
The agent detects your environment and handles everything automatically.
Without an agent (manual)
1. Run setup:
npx @httpayer/mcp setup
Get your API key at app.httpayer.com when prompted.
Flags:
| Flag | Description |
|---|---|
--key sk-live-... |
Provide key non-interactively |
--client <name> |
Target client: claude-code, claude-desktop, cursor, windsurf, opencode, zed, cline, warp, codex |
--scope user|project |
Claude Code scope (default: user) |
--yes / -y |
Skip all prompts |
--update-key |
Replace existing key |
2. Add to your client:
Claude Code:
claude mcp add httpayer --scope user -- npx -y @httpayer/mcp@latest
Claude Desktop (~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows):
{
"mcpServers": {
"httpayer": {
"command": "npx",
"args": ["-y", "@httpayer/mcp@latest"]
}
}
}
Cursor (.cursor/mcp.json), Windsurf (.windsurf/mcp.json), Cline (.cline/mcp_settings.json):
{
"mcpServers": {
"httpayer": {
"command": "npx",
"args": ["-y", "@httpayer/mcp@latest"]
}
}
}
OpenCode (opencode.json or ~/.config/opencode/config.json):
{
"mcp": {
"httpayer": {
"type": "local",
"command": ["npx", "-y", "@httpayer/mcp@latest"],
"enabled": true
}
}
}
Zed:
{
"context_servers": {
"httpayer": {
"command": {
"path": "npx",
"args": ["-y", "@httpayer/mcp@latest"]
}
}
}
}
3. Restart your client and verify:
Ask your agent: "fetch https://api.httpayer.com/demo/v1/base-weather"
A weather response means HTTPayer is working.
How it works
User prompt
│
▼
AI agent (Claude Code, Cursor, Windsurf...)
│ uses MCP tools
▼
@httpayer/mcp (local MCP server via npx)
│ REST calls with x-api-key header
▼
api.httpayer.com
│ proxy handles x402 payment to target
▼
Target x402-gated API
Runtime flow
- Your client launches the MCP server via
npx -y @httpayer/mcp@lateston startup (stdio transport). - The server reads the API key from
~/.httpayer/mcp-config.json. - The agent receives the tool list and system instructions in its context.
- When the agent calls
fetch, the MCP server forwards the request toPOST https://api.httpayer.com/proxy. - HTTPayer's proxy detects a 402, pays using your credits, retries, and returns the final response.
- The result (status, body, headers) comes back to the agent.
MCP tools reference
get_balance
Check credit balance and daily usage.
Input: none
Example response:
{
"account_id": "account_123",
"mainnet": {
"credits_balance": 50000,
"daily_limit": 100000,
"daily_spend": 15500,
"daily_remaining": 84500
}
}
fetch
Make an HTTP request to any x402-enabled endpoint. Payment is handled automatically.
Input:
| Field | Type | Required | Description |
|---|---|---|---|
url |
string | yes | Target URL |
method |
string | no | GET, POST, PUT, DELETE, PATCH — default GET |
body |
object | no | JSON request body |
params |
object | no | Query string parameters |
headers |
object | no | Additional request headers |
timeout |
number | no | Timeout in seconds, max 120 |
Example response:
{
"status": 200,
"body": { "data": "..." },
"headers": { "content-type": "application/json" }
}
On 502, the response includes webhook_id for async polling.
simulate
Dry-run a fetch. Returns cost estimate without spending credits.
Input: Same as fetch (except timeout).
Example response:
{
"requiresPayment": true,
"proxyFeeBreakdown": {
"targetAmount": 0.01,
"proxyFee": 0.0003,
"totalCreditsCharged": 10.3
}
}
get_topup_link
Returns the dashboard URL to add credits. Show to user when balance is low.
Input: none
check_limits
Check global HTTPayer system daily limits and remaining capacity.
Input: none
get_webhook_status
Poll the status of an async operation. Use when fetch returns a 502 with webhook_id.
Input: webhook_id (string, required)
Status values: pending, success, success_refunded, payment_failed, upstream_error, internal_error, rate_limited
HTTPayer API reference
Authentication: x-api-key: sk-live-... header on all requests.
| Method | Path | Tool |
|---|---|---|
GET |
/v1/credits/balance |
get_balance |
POST |
/proxy |
fetch |
POST |
/proxy/sim |
simulate |
GET |
/limits |
check_limits |
GET |
/webhooks/{id} |
get_webhook_status |
Proxy endpoint
POST https://api.httpayer.com/proxy
{
"api_url": "https://target.example.com/endpoint",
"method": "GET",
"json": { "key": "value" },
"params": { "query": "param" },
"headers": { "Custom-Header": "value" },
"timeout": 30
}
Only api_url and method are required.
Status codes:
| Code | Meaning |
|---|---|
200 |
Success |
402 |
Insufficient credits |
429 |
Rate limited |
500 |
Proxy error |
502 |
Target refused payment — includes webhook_id |
Configuration
API key stored at: ~/.httpayer/mcp-config.json
{ "apiKey": "sk-live-..." }
To update: npx @httpayer/mcp setup --update-key
x402 protocol overview
x402 is an HTTP-native micropayment protocol using the 402 Payment Required status code.
Without HTTPayer:
- Client hits endpoint → gets
402+ payment requirements - Client pays on-chain (requires wallet + USDC)
- Client retries with payment proof
With HTTPayer:
- Client calls
POST /proxy { api_url, method, ... } - HTTPayer detects
402, pays using your credits - HTTPayer retries and returns the final response
All blockchain interaction happens on HTTPayer's side.
Credit system
| Unit | Value |
|---|---|
| 1 credit | 0.001 USDC |
| 1 USDC | 1,000 credits |
| Proxy fee | 3% of target payment |
Top up at app.httpayer.com. Below 100 credits, the agent will prompt you to top up.
Error handling
Setup errors
| Situation | Behavior |
|---|---|
| Key format invalid | Print error, exit 1 |
| Key rejected (401) | Print "API key rejected", exit 1 |
| Network unreachable | Print reason, exit 1 |
MCP tool errors
All errors return isError: true — the server stays alive and the agent gets a readable message.
| Situation | Message |
|---|---|
| No config | "No HTTPayer API key configured. Run: npx @httpayer/mcp setup" |
| API non-2xx | "HTTPayer {status}: {body}" |
| Unknown tool | "Unknown tool: {name}" |
© 2026 HTTPayer Inc.
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。