Serena
Turns an LLM into a coding agent that works directly on your codebase by providing semantic retrieval and editing capabilities.
About
Serena is a powerful coding agent toolkit designed to transform Large Language Models (LLMs) into fully-fledged coding agents capable of operating directly on a codebase. It provides essential semantic code retrieval and editing tools similar to those found in an IDE, extracting code entities at the symbol level and exploiting relational structure. Serena supports integration with LLMs through the Model Context Protocol (MCP), offering an MCP server compatible with Claude Desktop, IDEs like VSCode, and soon ChatGPT. It also integrates with Agno, a model-agnostic agent framework, enabling the use of various LLMs from providers like Google, OpenAI, Anthropic, Ollama, Together, and Anyscale. Serena enhances the capabilities of LLMs with free and open-source tools for semantic code analysis and manipulation.
Key Features
- Integration with LLMs via MCP server
- Language server protocol (LSP) support for multiple programming languages
- Semantic code retrieval and editing capabilities
- Agno-based agent framework for model-agnostic LLM integration
- Compatibility with Claude Desktop, VSCode, and other IDEs
- 520 GitHub stars
Use Cases
- Autonomous execution of coding tasks from analysis to commit
- Semantic code analysis, planning, designing, and refactoring
- Integration with various LLMs using MCP or Agno