mcp-marketplace 0.1.0

Open MCP Marketplace, MCP Registry, MCP client, OneKey MCP Router, MCP Tool Search of commercial MCPs

MCP Marketplace Python SDK

MCP Marketplace Python Package is a common interface to give you access to public MCP Servers, Tools, Configurations. It supports various API endpoint (such as pulsemcp.com, deepnlp.org, etc).

PyPI|Document|MCP Marketplace|AI Agent Search|MCP Router Ranking List

Features

0. A Lightweight MCP Client (Like Cursor,Claude) and CLI mcpm for personal use and benchmarking tools/LLMs
1. Search API of MCP Tools: Users can search MCP Servers Meta Info and tools fit for mcp.json by query, such as "map", "payment", "browser use"
2. List MCP Tools API: And Allow LLM and AI Apps to Find Your MCP Server
3. Load MCP Config: Get latest mcp_config.json files
4. Development Customized API and Endpoint to Support Your Local MCP Marketplace API or Services
5. Registry: Allow Users to register the MCP Marketplace create, delete, update their MCP servers through various endpoints. (WIP)

0. Start MCP Client using mcpm CLI tool for Personal Use or Benchmark Purposes

Installation

You can install the command line mcpm tool from mcp-marketplace pypi package, and start a MCP Client using local mcp_config.json file

## Basic Usage MCP Index and Search
pip install mcp-marketplace

## MCPClient Supports User Defined mcp_config.json, Just Like Other Clients Cursor/Claude
pip install mcp-marketplace[mcp_tool_use] 

Install From GitHub Source

git clone https://github.com/aiagenta2z/mcp-marketplace
cd mcp-marketplace/python
pip install -e .

## go to different folder test mcpm cli
mcpm run

## Missing Dependency, Please report to use via issues

The CLI 'mcpm' will be in python path

mcpm run --host 0.0.0.0 --port 5000

## Use Local Config File, Go To GitHub Get the config https://github.com/aiagenta2z/mcp-marketplace

cd python/tests

mcpm run --port 5000 --config "./mcp_config_onekey.json"

mcpm run --port 5000 --config "./mcp_config.json"

Then you can visit http://0.0.0.0:5000 for web console and http://0.0.0.0:5000/mcp for mcp management

Benchmark

You can also run various MCP benchmark using the rest API From the Client

API	Description
/api/query	The endpoint which you can post the tool parameters to get the tool/call results

CURL Tests

Once you started the service and run the target mcp server, you can run the tools calls through the REST API.

Let's say you want to post to server: puppeteer and test tools: puppeteer_navigate to navigate chrome to a webpage.

Endpoint: http://127.0.0.1:5000/api/query

curl -X POST -H "Content-Type: application/json" -d '{
    "server_id": "puppeteer",
    "tool_name": "puppeteer_navigate",
    "tool_input": {
        "url": "https://arxiv.org/list/cs/new"
    }
}' http://127.0.0.1:5000/api/query

Result

{"success":true,"data":["Navigated to https://arxiv.org/list/cs/new"],"error":null}%

REST GET/POST Request By Python/Typescript/etc

For example, post request to MCP: amap-amap-sse, Tool: Get Weather and get local weather. You need to start mcp server 'amap-amap-sse' before running command line.

cd ./tests

python run_mcp_request.py

## define input_params
        input_params = {
            "server_id": "amap-amap-sse", 
            "tool_name": 'maps_weather',
            "tool_input": {
                "city": "乌鲁木齐"
            }
        }

Result

{'success': True, 'data': ['{"city":"乌鲁木齐市","forecasts":[{"date":"2025-06-30","week":"1","dayweather":"多云","nightweather":"多云","daytemp":"31","nighttemp":"20","daywind":"西北","nightwind":"西北","daypower":"1-3","nightpower":"1-3","daytemp_float":"31.0","nighttemp_float":"20.0"},{"date":"2025-07-01","week":"2","dayweather":"多云","nightweather":"多云","daytemp":"28","nighttemp":"20","daywind":"西北","nightwind":"西北","daypower":"1-3","nightpower":"1-3","daytemp_float":"28.0","nighttemp_float":"20.0"},{"date":"2025-07-02","week":"3","dayweather":"多云","nightweather":"多云","daytemp":"28","nighttemp":"20","daywind":"西北","nightwind":"西北","daypower":"1-3","nightpower":"1-3","daytemp_float":"28.0","nighttemp_float":"20.0"},{"date":"2025-07-03","week":"4","dayweather":"多云","nightweather":"晴","daytemp":"30","nighttemp":"21","daywind":"西北","nightwind":"西北","daypower":"1-3","nightpower":"1-3","daytemp_float":"30.0","nighttemp_float":"21.0"}]}'], 'error': None}

Install

pip install mcp-marketplace

Dependency

pip install uvicorn mcp fastapi pydantic dotenv asyncio httpx mcp_marketplace uuid aiofiles logger anthropic jinja2

1. Search API of MCP Tools

Usage

Example 1. Search MCP Marketplace By Query or Server ID

The server id follows the same in the github repo ${owner}/${repo}

import mcp_marketplace as mcpm

## endpoint: deepnlp
mcpm.set_endpoint("deepnlp")
result = mcpm.search(query="map", page_id=0, count_per_page=20, mode="dict")

print ("DEBUG: run_setup_config_deepnlp result:")
print (result)


## endpoint: pulsemcp
mcpm.set_endpoint("pulsemcp")
result2 = mcpm.search(query="map", count_per_page=20, offset=0)

print ("DEBUG: run_setup_config_pulsemcp result:")
print (result2)

MCP Result

{
  "q": "map",
  "limit": 50,
  "items": [
    {
        "id": "",
        "content_name": "Google Maps",
        "publisher_id": "pub-google-maps",
        "website": "https://github.com/modelcontextprotocol/servers/tree/main/src/google-maps",
        "review_cnt": "1",
        "subfield": "MAP",
        "field": "MCP SERVER",
        "rating": "5.0",
        "description": "",
        "content_tag_list": "official",
        "thumbnail_picture": "http://118.190.154.215/scripts/img/ai_service_content/b7fe82a3ab985ce1a953f7b4ad9c5e01.jpeg"
    },    
  ]
}

2. List MCP Tools API

Example 2. List Tools of MCP Servers list_tools method

Let's choose the unique id of browser use mcp "/puppeteer/puppeteer". And we can search the MCP meta and list the tools as below.

    import mcp_marketplace as mcpm
        
    result_q = mcpm.search(query="browser use", mode="list", page_id=0, count_per_page=100, config_name="deepnlp")
    result_id = mcpm.search(id="/puppeteer/puppeteer", mode="list", page_id=0, count_per_page=100, config_name="deepnlp")
    tools = mcpm.list_tools(id="/puppeteer/puppeteer", config_name="deepnlp_tool")

    print (f'{result_id}')
    print (f'{tools}')

Example 2.1 List Batch Tools of MCP Servers list_tools_batch method

Let's choose the unique id of browser use mcp "/puppeteer/puppeteer". And we can search the MCP meta and list the tools as below.

    import mcp_marketplace as mcpm
        
    servers_ids = ["puppeteer/puppeteer", "google-maps/google-maps"]
    query_params_list = [{"id": server_id, "config_name": "deepnlp_tool"} for server_id in server_ids]
    batch_tools_result = mcpm.list_tools_batch(query_params_list)
    print (f'DEBUG: list_tools_batch servers_ids {servers_ids}')
    print (f'DEBUG: list_tools_batch batch_tools_result {batch_tools_result}')

3. Load MCP Config

Example 3.1 load_config_batch Method

    import mcp_marketplace as mcpm
    server_ids = ["puppeteer/puppeteer", "mendableai/firecrawl-mcp-server", "google-maps/google-maps"]
    mcp_config_result = mcpm.load_config_batch(server_ids, config_name="deepnlp_server")
    print (mcp_config_result)

4. Development Customized API and Endpoint

If you have MCP server APIs and want to customized the mcp_marketplace lib. You can pass the get_customized_url() function to the lib

The get_customized_url method assemble the endpoint and your id in a customized way and will post request to your Endpoint.

def customized_client():


    def get_customized_url(**param):
        id_value = param["id"] if "id" in param else ""
        base_url = param["endpoint"] if "endpoint" in param else ""
        return base_url + "/" + id_value
    
    mcpm.set_endpoint("http://www.deepnlp.org/api/mcp_marketplace/v1/tools")
    mcpm.get_customized_endpoint = get_customized_url
    
    ## **param: {"id": "puppeteer/puppeteer", "endpoint":"your_endpoint"}
    tools = mcpm.list_tools(id="puppeteer/puppeteer", endpoint=mcpm.endpoint)

5. Registry

API Configuration

Support Endpoint

Endpoint	API	Document
http://www.deepnlp.org/api/mcp_marketplace/v1	http://www.deepnlp.org/api/mcp_marketplace/v1?mode=list&query=map&page_id=0&count_per_page=100	-
https://api.pulsemcp.com/v0beta/servers	https://api.pulsemcp.com/v0beta/servers?query=image&count_per_page=10	https://www.pulsemcp.com/api

deepnlp.org

# list mode return paginated list
http://www.deepnlp.org/api/mcp_marketplace/v1?mode=list&query=map&page_id=0

http://www.deepnlp.org/api/mcp_marketplace/v1?mode=list&query=map&page_id=0&count_per_page=100

http://www.deepnlp.org/api/mcp_marketplace/v1?mode=list&query=map&offset=50&count_per_page=5


# dict mode return dict group by category

http://www.deepnlp.org/api/mcp_marketplace/v1?mode=dict&query=map&page_id=0

http://www.deepnlp.org/api/mcp_marketplace/v1?mode=dict&query=map&page_id=0&count_per_page=5

http://www.deepnlp.org/api/mcp_marketplace/v1?mode=dict&query=map&offset=50&count_per_page=5


# List Tools

http://www.deepnlp.org/api/mcp_marketplace/v1/tools/puppeteer/puppeteer

Parameter

param	type	example
mode	string	"list", "dict", different use scenario
query	string	e.g. query="Image"
page_id	integer	starting from 0
count_per_page	integer	5
offset	integer	Optional, Equivalent to (page_id * count_per_page) e.g. 0

pulsemcp.com

## query API
https://api.pulsemcp.com/v0beta/servers?query=image&count_per_page=10

## List API

https://api.pulsemcp.com/v0beta/servers

Parameter

param	type	example
query	string	e.g. query="Image"
count_per_page	integer	5
offset	integer	Equivalent to (page_id * count_per_page) e.g. 0

Related

• MCP Marketplace DeepNLP
• MCP Marketplace PulseMCP
• Pypi
• Github
• AI Agent Marketplace