"model": "GLM-4-Flash", "api_url": "https://open.bigmodel.cn/api/paas/v4", "api_key": "", "other_params": { "temperature": 0.7, "max_tokens": 4096 } } ``` Or to use Anthropic's Claude model via LiteLLM: ```JSON "llm": { "provider": "litellm", "model": "anthropic/claude-3-opus-20240229", "api_key": "", "other_params": { "temperature": 0.7, "max_tokens": 4096 } } ``` ## Example ### Ex 1: ![example](./docs/images/mcp-client-weather.png) ### Ex 2: ![example](./docs/images/simple-mcp-client-ex2.png) ## Installation There are several ways to install the client: ### Using UV (recommended) Install UV ```bash curl -LsSf https://astral.sh/uv/install.sh | sh ``` ```bash uv venv source .venv/bin/activate python -m build ``` ### Using pip with requirements.txt ```bash python -m venv .venv source .venv/bin/activate pip install -r requirements.txt pip install -e . ``` ### Using pip directly ```bash pip install -e . ``` ### Using pip with pyproject.toml ```bash pip install . ``` ## Configuration The client uses a configuration file to define LLM providers and MCP servers. By default, it looks for `config.json` in the current directory, or it creates one with default values. Example configuration: ```json { "llm": { "provider": "openai", "model": "claude-3-7-sonnet-20250219", "api_url": "https://api.gptoai.top/v1", "api_key": "", "other_params": { "temperature": 0.7, "max_tokens": 4096 } }, "mcpServers": { "k8s": { "type": "sse", "url": "http://192.168.182.128:8000/sse", "command": null, "args": [], "env": {} } }, "default_server": "k8s" } ``` ## Usage Start the client: ```bash simple-mcp-client ``` Or run it directly: ```bash python -m simple_mcp_client.main ``` ## Available Commands - `help`: Show help message - `connect `: Connect to an MCP server - `disconnect `: Disconnect from an MCP server - `servers`: List available MCP servers - `tools [server_name]`: List available tools, optionally from a specific server - `resources [server_name]`: List available resources, optionally from a specific server - `prompts [server_name]`: List available prompts, optionally from a specific server - `formats [server_name]`: List available prompt formats, optionally from a specific server - `execute [arg1=val1 ...]`: Execute a specific tool with arguments - `get-resource [server_name] `: Get a resource from an MCP server - `get-prompt [server_name] [format=] [arg1=val1 ...]`: Get a prompt from an MCP server - `chat`: Start a chat session with the configured LLM and active MCP tools - `config show`: Show current configuration - `config llm [model=] [api_url=] [api_key=] [param=value ...]`: Configure LLM provider - `reload`: Reload configuration from file - `exit`: Exit the program ## Command Examples ### Resources Command The `resources` command lists available resources and resource templates from connected MCP servers. List all resources from all connected servers: ```bash resources ``` List resources from a specific server: ```bash resources k8s ``` Output example: ``` Resources from k8s - Direct Resources ┏━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━┓ ┃ URI ┃ Name ┃ MIME Type ┃ ┡━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━┩ │ k8s://namespaces │ Kubernetes Namespaces │ application/json │ │ k8s://resources │ Kubernetes Resources │ application/json │ └──────────────────┴───────────────────────┴

Simple MCP Client

Comments