MCP Testing

https://github.com/yliust/mcp-testing
Admin
This project provides documentation and examples for testing and improving the descriptions of tools...
Tokens:1,055
Snippets:8
Trust Score:3.2
Update:1 year ago
Show doc for...
Tokens:
Raw
### Example of Improved Tool Description with Query Example

Source: https://github.com/yliust/mcp-testing/blob/master/README.md

This Python function demonstrates how to refine a tool's description by embedding a clear query example within its docstring. The 'search_pubchem_by_name' function's docstring now includes details on arguments, return types, and a concrete 'Query example' to guide users.

```python
def search_pubchem_by_name():
    """
    Search PubChem for compounds matching a chemical name.

    Args:
        name: The chemical name to search for
        **kwargs: Additional parameters for the API request
                    (e.g., max_records: int)
    
    Query example: {
        "name": "aspirin",
        "kwargs": {}
    }

    Returns:
        Dictionary containing search results with compounds that match the name
    """
    pass
```

--------------------------------

### Install Python Dependencies for Langchain MCP Adapters

Source: https://github.com/yliust/mcp-testing/blob/master/README.md

This snippet shows how to install the necessary Python package 'langchain_mcp_adapters' using pip, which is required to interact with Multi-Server MCP clients.

```bash
pip install langchain_mcp_adapters
```

--------------------------------

### Invoke PubChem Search Tool with Example Query

Source: https://github.com/yliust/mcp-testing/blob/master/README.md

This Python code demonstrates how to invoke the 'search_pubchem_by_name' tool with a sample query for 'aspirin'. It shows how to structure the 'query_args' dictionary and use 'tool.ainvoke' to execute the tool, verifying its functionality based on its clear description.

```python
query_args = {
    "name": "aspirin", # Name of the compound
    "kwargs": {} # other parameters, can be set to an empty dict
}

result = await tool.ainvoke(query_args)
print(result)
```

--------------------------------

### Example of Clear Tool Description Retrieval

Source: https://github.com/yliust/mcp-testing/blob/master/README.md

This Python snippet illustrates how to retrieve and display the description and argument schema for a tool ('gsea_search') that has a clear and comprehensive description. The detailed output (not shown in code) typically includes the tool's function, query format, example usage, and explicit arguments, making it easy to understand and use.

```python
tool_name = 'gsea_search'
tool = tool_map.get(tool_name)
show_tool_description(tool)
```

--------------------------------

### Example of Unclear Tool Description Retrieval

Source: https://github.com/yliust/mcp-testing/blob/master/README.md

This Python snippet demonstrates how to retrieve and display the description and argument schema for a tool ('get_activity') that has an unclear description. The accompanying output (not shown in code) typically reveals a vague description and an empty argument schema, making it difficult to understand the tool's purpose or usage.

```python
tool_name = 'get_activity'
tool = tool_map.get(tool_name)
show_tool_description(tool)
```

--------------------------------

### Connect to MCP Client and Retrieve Tool Descriptions

Source: https://github.com/yliust/mcp-testing/blob/master/README.md

This Python code demonstrates how to connect to a Multi-Server MCP client, retrieve a list of available tools, and then display the description and argument schema for a specific tool (e.g., 'get_cell_line_by_id'). It highlights how to access 'tool.description' and 'tool.args_schema'.

```python
import asyncio
from langchain_mcp_adapters.client import MultiServerMCPClient

mcp_servers = {}

# 连接mcp服务
client = MultiServerMCPClient(mcp_servers)
tools = await client.get_tools()
print(f"✅ 成功连接！发现 {len(tools)} 个工具:")

# 获取工具名称到工具的映射
tool_map = {tool.name: tool for tool in tools}

# 指定tool的name
tool_name = 'get_cell_line_by_id'
tool = tool_map.get(tool_name)
assert tool is not None, f"Tool {tool_name} not found!"

# 获取tool的description等信息
def show_tool_description(tool):
    print(f'Tool Name: {tool.name}\n')
    print('=' * 5, ' Tool Description ', '=' * 5)
    print(tool.description, '\n')
    print('=' * 5, ' Input Args Schema ', '=' * 5)
    print(tool.args_schema, '\n')

show_tool_description(tool)
```

--------------------------------

### Display Description for PubChem Search Tool

Source: https://github.com/yliust/mcp-testing/blob/master/README.md

This Python snippet retrieves and displays the description and argument schema for the 'search_pubchem_by_name' tool from the PubChem database. It serves as a prerequisite for demonstrating how to test a clearly described tool.

```python
tool_name = 'search_pubchem_by_name'
tool = tool_map.get(tool_name)
show_tool_description(tool)
```

--------------------------------

### External API Documentation Links for MCP Tools

Source: https://github.com/yliust/mcp-testing/blob/master/README.md

This section provides direct links to the original API documentation for various Multi-Server MCP tool packages, including Chembl, KEGG, PubChem, and String. These links are crucial for understanding tools with unclear internal descriptions.

```APIDOC
chembl_mcp API Documentation: https://www.ebi.ac.uk/chembl/api/data/docs
kegg_mcp API Documentation: https://rest.kegg.jp
pubchem_mcp API Documentation: https://pubchemdocs.ncbi.nlm.nih.gov/programmatic-access
string_mcp API Documentation: https://version-12-0.string-db.org/cgi/help?sessionId=bJAGYvuWb0TA
```
Tokens:
Raw