mcpField
M
Mcp Api Client

Mcp Api Client

Local
github.com/molavecDeveloper Tools7Updated 69w ago

About Mcp Api Client

Mcp Api Client is an MCP server that turns YAML endpoint definitions into MCP tools and fetch APIs for calling external APIs. It is intended for integrations and prototypes that need API access configured without writing custom client code.

Key features

  1. 01Define multiple endpoints in YAML
  2. 02Generate MCP tools from configuration
  3. 03Generate MCP fetch APIs
  4. 04Support HTTP methods including GET and POST
  5. 05Detect path and query parameters
  6. 06Configure API headers

Use cases

  1. 01Connect an AI model to external APIs
  2. 02Expose automation webhooks through MCP
  3. 03Prototype API workflows with YAML
  4. 04Present multiple APIs through one MCP interface

MCP support

The server reads YAML API definitions and exposes configured API calls as MCP tools and fetch APIs to MCP clients.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"api-client": {
"command": "npx",
"args": [
"-y",
"mcp-api-client"
]
}
}
}
Run commandnpx -y mcp-api-client

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmmcp-api-client

Tools

7

Saved tool list · updated 2w ago

This tool takes no parameters.

Details

From the project README

MCP API Client (Beta)

A Model Context Protocol (MCP) server that call APIs using YAML config files.

This server allows you connect to APIs using simple a config YAML file! 🤩

Ideal for: #lazy-programmers #nocode #lowcode #ai-integrators #automation #api-gateway #data-pipelines #rapid-prototyping #llm-integration #makers

If you don't have advance APIs call skills using a programming language or you just want a fast way to connect your AI model with an API using MCP, you will LOVE 💖 this MCP.

🎬▶️ Example Video😉✨

Check this video using included Test Server and Claude

Support with your Click!

Make a reaction and give me a motivational comment on mcp.so repository

Help me on mcp.so

... and yes! A donation to help cover my coffee and rent would be amazing! 🙏🙏

ko-fi

Features
  • Configure multiple API endpoints using YAML
  • Automatic MCP tool creation from YAML config file.
  • Automatic MCP fetch API creation from YAML config file.
  • Supports all HTTP methods (GET, POST, PATCH, PUT, DELETE) [✨PARTIAL✨]
    • [✅] GET API without parameters
    • [✅] GET API with query parameters
    • [✅] GET API with path parameters
    • [✅] POST API
    • [✅] PATCH API with path parameters
    • [✅] PUT API with
    • [✅] DELETE API with path parameters
  • Set API Headers [✨WORKING✨]
  • Set API Token [✨ 🫣🔴COMMING SOON✨]
  • Environment variable interpolation for sensitive data [✨ 🫣🔴COMMING SOON✨]
  • Automatic PATH parameter detection from URL (like /users/{id})[✨WORKING✨]
  • Automatic QUERY parameter detection from URL (like /users?page={page}&limit={limit}) [✨WORKING✨]
  • Reduce Prompt tokens used setting up certain APIs for your project
  • ...and best of all, SAVE COUNTLESS HOURS FOR HUMANITY 👏😎😉 by connecting your LLM to any API.
Use cases
  • Connect to N8N, Make.com or Zapier webhook: Make actions in other platforms connecting to automations or integration tools.
  • AI Model Integration: Quickly connect LLMs or other AI models to external APIs for data retrieval, enrichment, or action execution without custom code.
  • Rapid Prototyping: Build and test API workflows using YAML configuration, ideal for hackathons, demos, or proof-of-concept projects.
  • No-Code/Low-Code Automation: Enable non-developers to define and call APIs securely using simple YAML files and environment variables.
  • Data Pipelines: Orchestrate API calls as part of ETL or data processing pipelines, leveraging MCP tools for modularity and reuse.
  • API Gateway: Expose multiple backend APIs through a unified MCP interface, simplifying integration for downstream consumers.
  • Secure API Access: Manage secrets and tokens via environment variables, keeping sensitive data out of source code and YAML files.

Getting started

  1. Get a copy of default api.yaml.
npx -y mcp-api-client --init
  1. You can test default api.yaml with express test server test/server.js. Execute in other terminal.
npx -y mcp-api-client --test-server
  1. Add MCP to your agent config file.
  • Example for Claude Client settings file:
"mcp": {
    "servers": {
      "mcp-api-client": {
          "command": "npx",
          "args": [
              "-y",
              "mcp-api-client",
              "path-to-api.yaml"
          ]
      },
    }
  }

  • Example for Visual Studio Code settings file:
"mcp": {
    "servers": {
      "mcp-api-client": {
          "command": "npx",
          "args": [
              "-y",
              "mcp-api-client",
              "path-to-api.yaml"
          ]
      },
    }
  }

  1. Just Ask!

Ask to Claude LLM in MCP client

YAML Configuration File

Each API YAML file can include global metadata with MCP information:

metadata: General information about the API set

and a list of API endpoint definitions:

apis: A list of API endpoint definitions. Each endpoint should have
Example API YAML config file example
metadata:
  name: API Examples
  version: 1.0.0
  description: This file contains examples of various API endpoints for testing purposes.

apis:
  - name: getUser
    url: https://api.example.com/users/{id}
    method: GET
    content:
      headers:
        Accept: application/json
  
  - name: createUser
    url: https://api.example.com/users
    method: POST
    api-token: ${API_TOKEN}
    content:
      headers:
        Content-Type: application/json
      body:
        name: string
        email: string
  ...

See public/apis.yaml for a complete example with metadata and all HTTP methods and parameter types.

API_TOKEN from .env (Not tested yet!)

Create a .env file for your API tokens:

API_TOKEN=your_api_token_here

Bugs and Requests

https://github.com/molavec/mcp-api-client/issues

Contributing

  1. Fork the repository
  2. Create your feature branch
  3. Commit your changes
  4. Push to the branch
  5. Create a new Pull Request

License

MIT

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.