MCP server with structured http_request tool for AI agents. Built for Claude Code and any MCP-compatible client. Single static Go binary, zero external dependencies at runtime.
rest-api-mcp
A lightweight MCP (Model Context Protocol) server that provides a structured http_request tool for AI agents. Eliminates the need for fragile curl commands that break on Windows due to quoting/escaping issues.
Why?
AI agents (like Claude Code) frequently need to make HTTP requests during API development. Using curl via Bash is unreliable on Windows — quoting JSON bodies, escaping special characters, and handling headers all break differently across shells. This MCP server provides a single, structured http_request tool that works identically on all platforms.
Quick Start
Register in Claude Code
The register subcommand automatically adds the server to Claude Code's config — no manual JSON editing needed.
# Register for current project (creates .mcp.json)
rest-api-mcp register project
# Register with a base URL for API development
rest-api-mcp register project . -- --base-url http://localhost:8080
# Register globally for all projects (writes to ~/.claude.json)
rest-api-mcp register user
Arguments after -- are forwarded to the MCP server on every startup.
More examples
# Authenticated API with default headers
rest-api-mcp register project . -- \
--base-url https://api.example.com \
--default-header "Authorization: Bearer YOUR_TOKEN_HERE" \
--default-header "Content-Type: application/json"
# Corporate proxy / self-signed certs
rest-api-mcp register project . -- \
--base-url https://internal-api.corp.local \
--proxy http://proxy.corp.local:8080 \
--insecure
Manual configuration
You can also edit the config files directly. The register command generates entries like this in .mcp.json or ~/.claude.json:
{
"mcpServers": {
"rest-api": {
"command": "/path/to/rest-api-mcp",
"args": ["--base-url", "http://localhost:8080"]
}
}
}
Full configuration example with all options
{
"mcpServers": {
"rest-api": {
"command": "/path/to/rest-api-mcp",
"args": [
"--base-url", "http://localhost:3000",
"--default-header", "Authorization: Bearer YOUR_TOKEN_HERE",
"--default-header", "Content-Type: application/json",
"--default-header", "Accept: application/json",
"--timeout", "15s",
"--max-response-size", "102400",
"--retry", "2",
"--retry-delay", "500ms"
]
}
}
}
Installation
Download binary
Pre-built binaries for Windows, macOS, and Linux are available on the Releases page.
Build from source
Requires Go 1.22+.
git clone https://github.com/lexandro/rest-api-mcp.git
cd rest-api-mcp
go build -o rest-api-mcp.exe .
CLI Flags
| Flag | Default | Description |
|------|---------|-------------|
| --base-url | (none) | Base URL prepended to relative paths |
| --default-header | (none) | Default header (repeatable), format: Key: Value |
| --timeout | 30s | Default request timeout |
| --max-response-size | 51200 | Maximum response body size in bytes (default 50KB) |
| --proxy | (none) | HTTP/HTTPS proxy URL |
| --retry | 0 | Number of retry attempts for failed requests |
| --retry-delay | 1s | Delay between retries |
| --insecure | false | Skip TLS certificate verification |
Tool: http_request
A single, versatile tool for making HTTP requests.
Parameters
| Parameter | Type | Required | Description |
|-----------|------|----------|-------------|
| method | string | yes | HTTP method: GET, POST, PUT, DELETE, PATCH, HEAD, OPTIONS |
| url | string | yes | Full URL or relative path (if --base-url is set) |
| headers | object | no | Request headers as key-value pairs |
| body | string | no | Request body (typically JSON) |
| queryParams | object | no | Query parameters as key-value pairs |
| timeout | string | no | Per-request timeout override (e.g., 10s, 500ms) |
| followRedirects | boolean | no | Follow HTTP redirects (default: true) |
| includeResponseHeaders | boolean | no | Include response headers in output (default: false) |
Response Format
Compact, token-efficient output designed to minimize context window usage:
HTTP 200 OK (154ms)
{"id": 1, "name": "example"}
With includeResponseHeaders: true:
HTTP 200 OK (154ms)
Content-Type: application/json
X-Request-Id: abc123
{"id": 1, "name": "example"}
Large responses are automatically truncated:
HTTP 200 OK (1.2s)
{"data": [...first 50KB...]}
[truncated, showing 51200 of 245891 bytes]
Token Efficiency
- No response headers by default — saves ~200-500 tokens per request
- 50KB response limit — prevents dumping huge payloads into context
- Compact status line —
HTTP 200 OK (154ms)instead of verbose curl output - No request echo — the agent already knows what it sent
- Error as text —
Request failed: connection refusednot a stack trace
Examples
Simple GET
{ "method": "GET", "url": "/api/users" }
POST with JSON body
{
"method": "POST",
"url": "/api/users",
"headers": { "Content-Type": "application/json" },
"body": "{\"name\": \"John\", \"email\": \"john@example.com\"}"
}
GET with query parameters
{
"method": "GET",
"url": "/api/search",
"queryParams": { "q": "hello", "limit": "10" }
}
PUT with timeout override
{
"method": "PUT",
"url": "/api/users/123",
"body": "{\"name\": \"Updated\"}",
"timeout": "5s"
}
GET without following redirects
{
"method": "GET",
"url": "/api/short-link",
"followRedirects": false,
"includeResponseHeaders": true
}
Development
Build
go build -o rest-api-mcp.exe .
Test
go test ./...
Vet
go vet ./...
License
MIT