MCPFast / Tools / Sivru: Code Search and Observability for AI Agents
Sivru provides hybrid code search (BM25+semantic) and observability for AI agents, with optional reranking and local self-benchmarking.
View on GitHub→Sivru is a powerful tool designed to enhance the development and debugging of AI agents. It offers a robust solution for code search and observability, integrating advanced techniques to provide developers with deeper insights into their agent's behavior and performance. By combining traditional search methods with semantic understanding, Sivru aims to streamline the process of finding relevant code and monitoring agent execution.
Sivru provides a hybrid approach to code search, leveraging both BM25 (a keyword-based retrieval model) and semantic search capabilities. This allows developers to find code snippets based on exact matches as well as conceptual relevance. Beyond search, Sivru offers comprehensive observability features for AI agents. This includes the ability to track agent execution, identify potential issues, and analyze performance metrics. The tool also supports optional reranking of search results to prioritize the most relevant information and includes local self-benchmarking for performance evaluation.
Sivru is specifically targeted at AI developers and engineers working with AI agents. If you are building, debugging, or optimizing AI agents, Sivru can significantly improve your workflow. It is particularly useful for those who need to quickly find specific code related to agent logic, understand how their agents are behaving in real-time, and benchmark their performance effectively. Developers looking for advanced code search and deep observability into their AI agent projects will find Sivru to be an invaluable asset.