johnrobinsn/askit

AskIt is a flexible asyncio Python library and CLI tool that allows various LLM models to extend their abilities by invoking services from Model Context Protocol (MCP) servers and by calling locally-defined Python functions.
## Features
Simple and Lightweight
Connect to multiple MCP servers simultaneously
Supports STDIO, SSE and Streamable-HTTP MCP Servers
Support for Tool Use/Function Calling with MCP Servers Written in any Language
Support for Tool Use/Function Calling with Locally-Defined Python Functions
Supports Multiple LLM Providers
OpenAI and XAI (Grok) Currently Supported
Anthropic, Ollama, and LMStudio Support __Coming Soon__
Optional Support for Streaming LLM Responses
Securely Store API Keys in Environment Variables
## Installation
To get started with AskIt MCP, you need to have Python 3.11+ installed on your system. A python version manager is recommended. You can then install the package using pip:
bash
pip install git+https://github.com/johnrobinsn/askit.git
## Getting Started with the AskIt CLI
The easiest way to get started with AskIt is to use its integrated command-line interface (CLI) tool. The CLI allows you to interact with MCP servers and LLMs using natural language queries, making it easy to access and manipulate data from various sources.
The following command will leverage the default OpenAI model (gpt-4o-mini) to enter into a "chat-eval" loop with the configured LLM.
You just need to provide your OpenAI API key as a command line argument.
eg.
bash
python -m askit --api_key="your_openai_api_key_here"
You can also provide the OpenAI API key as a provider specific environment variable.
bash
OPENAI_API_KEY="your_openai_api_key_here" python -m askit
You can specify a different model or provider using command-line options. Here is an example using the XAI provider (Grok) with a specific model. You will need to provide your XAI API key as an environment variable or command line argument.
bash
python -m askit --model="grok-4-latest" --provider="XAI" --api_key="your_xai_api_key_here"
Find out more about Provider Specific Environment Variables.
## Connect MCP Servers to Askit
You can connect to multiple MCP servers by creating a mcp_config.json file in your current working directory. This file should contain the configuration for the MCP servers you want to connect to. The AskIt CLI will automatically load this configuration and connect to the specified MCP servers.
eg.
json
"mcpServers": {
"example-stdio": {
"command" : "node",
"cwd" : "/mnt2/code/mcpcalc",
"args": [
"mcpcalc.mjs"
],
"env": {
"debug_log": "false"
},
"example-streamable-http": {
"transport" : "http",
"url" : "http://127.0.0.1:8000/mcp",
"disabled" : false
As a reference, a sample mcp_config.json.example file is provided in the root of the repository. You can copy this file to mcp_config.json and modify it to suit your needs.
You can add as many MCP servers as you need and the client will connect to all of them and make their tools available to the configured LLM. There are two primary types of MCP servers supported by AskIt, stdio and Streamable HTTP. The former is used for MCP servers that are located on your local machine, while the latter can be used to access MCP servers that are hosted remotely.