mcpField
R
Run the MCP Client

Run the MCP Client

Local
github.com/dazzajiDeveloper Tools0Updated 97w ago

About Run the MCP Client

Run the MCP Client is a Python client for testing an already-running filesystem MCP server. It lists available tools, calls a selected tool, and prints results for connectivity and behavior checks.

Key features

  1. 01Connect to a filesystem MCP server
  2. 02List available tools
  3. 03Call tools with supplied arguments
  4. 04Print call results

Use cases

  1. 01Check client-to-server connectivity
  2. 02Test filesystem tools with chosen arguments
  3. 03Verify file creation, reading, or listing

MCP support

Connects to a running filesystem MCP server, discovers its available tools, and invokes a selected tool.

Configuration

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

No install method listed yet

We haven’t collected a package or remote endpoint for this server. Follow the instructions in the project README.

Repository

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

Run the MCP Client

Before running this client be sure to have the server up and running. For more info on that, see: https://github.com/dazzaji/filesystem

Step 1: Create the Virtual Environment and Install Dependencies Correctly

Use the following commands:

```bash
mkdir mcp-client-py
cd mcp-client-py
```

```bash
python -m venv venv
source venv/bin/activate
```

```bash
pip install -r requirements.txt
```

This creates a virtual environment, activates it, and then installs mcp and python-dotenv (and their dependencies) into that environment.

Step 2: Confirm or Correct the Environment Variables in .env

Create a .env file with your server path and the allowed directory. You can also specify tool and args for testing purposes.

   SERVER_PATH=/Users/dazzagreenwood/filesystem/dist/index.js
   ALLOWED_DIRECTORY=/Users/dazzagreenwood/mcp-hello/module1/files
   
  # Test with arguments:
   #TOOL=list_directory
   #ARGS='{"path": "/Users/dazzagreenwood/mcp-hello/module1/files", "recursive": true}'
   
   # Or test with other tools
   #TOOL=read_file
   #ARGS='{"path": "/Users/dazzagreenwood/mcp-hello/module1/files/test.txt"}'
Step 3: Running the Python Client

Run your client to test the server.

 python client.py
 ```
*   If you used the optional `TOOL` and `ARGS` add those arguments like so:
 ```bash
     python client.py --tool "list_directory" --args '{"path": "/Users/dazzagreenwood/mcp-hello/module1/files", "recursive": true}'
 ```
This will:

*   Connect to your filesystem server.
*   List available tools.
*  Call the `write_file` to create a test file with text content in the specified directory, if `tool` is not specified in the .env file.
*   Display the results to the console.

### Step 4: Verifying Results

After successfully running the client, check the following:

1.  **Client Output:** You should see output like `Available tools: [ ... ]`, and the result from the `write_file` (or other tool call).
2.  **File Creation:** Verify that `testfile.txt` has been created with the correct content inside `/Users/dazzagreenwood/mcp-hello/module1/files/`.  If testing with the read or list tools, make sure the output is what you'd expect.

### Testing Workflow

Here are steps to verify the project with the client and a server:

#### Starting the Server
1.  **Open a Terminal**. You will need to start the server from a terminal.
2.  **Navigate to Server Directory**. Change directory to the folder containing the server:
 ```
 cd /Users/dazzagreenwood/filesystem
 ```
3.  **Run the Server:**
 ```bash
 node dist/index.js "/Users/dazzagreenwood/mcp-hello/module1/files"
 ```
 *   Replace `/Users/dazzagreenwood/filesystem/dist/index.js` with the actual path to the compiled file.
 *   Replace `/Users/dazzagreenwood/mcp-hello/module1/files` with the allowed directory.

#### Testing with the Client

1.  **Open another Terminal window** or split your existing terminal tab.
2.  **Navigate to Client Directory** Change directories into the `mcp-client-py` folder:
  ```bash
  cd /Users/dazzagreenwood/mcp-client-server/mcp-client-py
  source .venv/bin/activate
  ```
3. **Run the Client in Default Mode:**
 To test basic connectivity and the `write_file` tool:
 ```bash
  python client.py
 ```
  This will attempt to call write file, to create a file named test.txt into the `/Users/dazzagreenwood/mcp-hello/module1/files` directory with the string `Hello from Python the new standup MCP client!` as it's content.

4.  **Run the client with specific tools and arguments**: To test list_directory, use the following:
```bash
   python client.py --server-path /Users/dazzagreenwood/filesystem/dist/index.js --allowed-dir /Users/dazzagreenwood/mcp-hello/module1/files --tool list_directory --args '{"path": "/Users/dazzagreenwood/mcp-hello/module1/files", "recursive": true}'
To test the `read_file` tool use:
 python client.py --server-path /Users/dazzagreenwood/filesystem/dist/index.js --allowed-dir /Users/dazzagreenwood/mcp-hello/module1/files --tool read_file --args '{"path": "/Users/dazzagreenwood/mcp-hello/module1/files/test.txt"}'
* Replace the paths in the examples above to correspond to your local paths.

If setup correctly you should see the output that confirms each of these actions.

By following these instructions, you can now test your MCP server and client with various tools using Python and TypeScript/Node.js!

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.