Tiny MCP Client

Example Use Cases

• 1

  Call external tool servers (e.g., weather, time, file save) and integrate the tool outputs into LLM responses via MCP.

• 2

  Prototype or test MCP workflows with both prompt-based and function-calling patterns, including streaming outputs.

• 3

  A configurable client to run single or multiple MCP servers (via config files or direct script paths) and experiment with resources and prompt templates.

Description

Tiny MCP Client is an open-source Python implementation of a Model Context Protocol (MCP) client that supports both prompt-based and function-calling workflows. It provides multiple entry points: a prompt-mode multi-server client, a function-calling single-server client, and a streaming function-calling client. The project makes it easy to discover server-provided tools/resources/prompts, call external MCP servers (stdio JS/Python), and route results back to the LLM for final responses. It's well suited for prototyping MCP integrations and testing with providers like DeepSeek or local deployments (Ollama/VLLM).