jiki

Name: jiki
Rating: 1.9 (17 reviews)
Author: teilomillet

by teilomillet·★ 17·Score 37

Jiki is a Python framework that connects LLMs to external tools via MCP, offering both orchestration and client capabilities.

ai-llmdeveloper-toolsproductivity

Forks

Open issues

13 mo ago

Last commit

2d ago

Indexed

Overview

Jiki is a comprehensive Python framework designed to bridge LLMs with external capabilities through the Model Context Protocol (MCP). It features an Orchestrator that manages conversation flow and an MCP Client for tool server communication. The framework supports multiple LLM backends via LiteLLM and offers flexible tool integration through auto-discovery or manual definition. Communication with tool servers is transparent, supporting both local scripts (stdio) and network services (HTTP/SSE), making it versatile for various application architectures.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

you:Building tool-augmented LLM applications with calculator or other custom tools

you:Creating interactive chat interfaces that can call external APIs via MCP

you:Integrating LLM capabilities with existing systems through standardized protocol

you:What transport methods does Jiki support for MCP communication?

you:How do I add custom tools to Jiki?

When to choose this

Choose Jiki if you're building Python applications that need to connect LLMs with external tools via MCP and want a flexible orchestration layer.

When NOT to choose this

Avoid Jiki if you need non-Python tool servers, as it currently only integrates with MCP servers in Python or requires custom implementation.

Comparable tools

fastmcpmcp-serverlangchain

Installation

Install the jiki package using your preferred package manager:

# Using pip
pip install jiki

# Or using uv (recommended for faster installation)
uv add jiki

Set Up API Key

Jiki uses [LiteLLM](https://litellm.ai/) internally, allowing it to work with a wide range of LLM providers (OpenAI, Anthropic, Gemini, etc.). You need to set the appropriate environment variable for your chosen provider.

# Example for Anthropic Claude (often used as default)
export ANTHROPIC_API_KEY=your_key_here

# Example for OpenAI
# export OPENAI_API_KEY=your_key_here

FAQ

What transport methods does Jiki support for MCP communication?: Jiki supports both stdio for local script communication and HTTP/SSE for network services, allowing flexible integration with MCP servers.
How do I add custom tools to Jiki?: You can either use auto-discovery to let Jiki discover tools from your MCP server, or provide tool definitions directly via a JSON file using the `tools` parameter.

Compare jiki with

jiki vs ultimate_mcp_server jiki vs mcp-server-chart jiki vs everything jiki vs filesystem jiki vs time

GitHub →

Last updated 2026-05-17 · Auto-generated from public README + GitHub signals.