sourcerykit

What it does

SourceryKit is a counterspell for hallucinating AI agents. This Python SDK intercepts and validates every tool call, API response, and MCP (Model Context Protocol) handoff to catch and prevent hallucinations before bad outputs propagate through your agent system. It provides runtime verification of agent outputs to ensure reliability and correctness.

How to set up

Install SourceryKit via pip: pip install sourcerykit. Import the SDK into your Python project and integrate it with your LangChain, LangGraph, or OpenAI agent workflows. Configure validation rules for tool calls and API responses according to your use case. The SDK hooks into your agent's execution pipeline to provide real-time output validation.

Key features

• Real-time validation of AI tool calls and outputs
• API response verification before propagation
• MCP handoff integrity checking
• Integration with LangChain, LangGraph, and OpenAI agents
• Hallucination detection and prevention
• Python SDK for seamless integration
• Prevents bad outputs from cascading through agent systems