MCP Client

Documentation for configuring the MCP client.

Note: This project uses uv in examples for speed and convenience, but uv is not required. You can use pip and standard Python commands instead. See Dependency Management for details.

uv run mcp-client --help
# usage: mcp-client [-h] [--sampling | --no-sampling] [--elicitation | --no-elicitation] [--logging | --no-logging] [--progress | --no-progress]
#                   [--debug | --no-debug] [--log_level {DEBUG,INFO,WARNING,ERROR,CRITICAL}] [--log_file {str,null}] [--language_model str]
#                   [--language_model_max_tokens int] [--language_model_temperature float] [--language_model_top_p float] [--language_model_timeout int]
#                   [--langfuse_enabled | --no-langfuse_enabled] [--langfuse_host {str,null}] [--langfuse_public_key {str,null}] [--langfuse_secret_key {str,null}]
#                   {azure_openai,hosted_openai,openai} ...
#
# Aggregate all configurations for the MCP client.
#
# options:
#   -h, --help            show this help message and exit
#   --sampling, --no-sampling
#                         (default: True)
#   --elicitation, --no-elicitation
#                         (default: True)
#   --logging, --no-logging
#                         (default: True)
#   --progress, --no-progress
#                         (default: True)
#   --debug, --no-debug   (default: False)
#   --log_level {DEBUG,INFO,WARNING,ERROR,CRITICAL}
#                         (default: WARNING)
#   --log_file {str,null}
#                         (default: mcp_client.log)
#   --language_model str  (default: gpt-4o-mini)
#   --language_model_max_tokens int
#                         (default: 4096)
#   --language_model_temperature float
#                         (default: 0.1)
#   --language_model_top_p float
#                         (default: 0.9)
#   --language_model_timeout int
#                         (default: 300)
#   --langfuse_enabled, --no-langfuse_enabled
#                         (default: False)
#   --langfuse_host {str,null}
#                         (default: null)
#   --langfuse_public_key {str,null}
#                         (default: null)
#   --langfuse_secret_key {str,null}
#                         (default: null)
#
# subcommands:
#   {azure_openai,hosted_openai,openai}
#     azure_openai
#     hosted_openai
#     openai

The subcommands further give configurations for different LLM providers. For example:

uv run mcp-client hosted_openai --help
# usage: mcp-client hosted_openai [-h] [--language_model_provider_type hosted_openai] [--hosted_openai_api_key str] [--hosted_openai_base_url str]
#                                 [--hosted_openai_headers {dict,null}]
#
# Define configurations for Hosted OpenAI.
#
# options:
#   -h, --help            show this help message and exit
#   --language_model_provider_type hosted_openai
#                         (default: hosted_openai)
#   --hosted_openai_api_key str
#                         (required)
#   --hosted_openai_base_url str
#                         (required)
#   --hosted_openai_headers {dict,null}
#                         (default: null)