Serena

Founded by Dominik Jain & Michael Panchenko in 2024

Give AI coding agents IDE-level code navigation and refactoring abilities

Cost

Free

Rating

★ People love it

Time to value

Quick Setup (< 1 hour)

You can use Serena to give AI coding agents semantic code tools similar to what an IDE provides — things like symbol-level navigation, cross-file renaming, reference lookups, and refactoring. It connects to any AI client via the Model Context Protocol (MCP), working with Claude Code, Codex, Copilot, Cursor, JetBrains, and others. Instead of relying on line numbers or text search, Serena operates at the symbol level, making AI-driven code edits faster and more reliable in large or complex codebases.

What Serena does

Rename symbols across all files in a codebase atomicallyFind all references to a function or class throughout a projectMove code modules between files and update all import referencesConnect the MCP server to an AI coding client via launch command or HTTP URLRun 20 routine coding tasks to benchmark agent performance with SerenaNavigate dependency relationships in a multi-language monorepoPerform safe refactors without manually tracking line numbersCross-file symbol renaming in a single call instead of multiple manual stepsReference lookups to find all usages of a symbol across the codebaseOperates at the symbol level rather than using line numbers or text patternsIntegrates with any MCP-compatible AI clientSupports multiple languages including multi-language monoreposCan run as an HTTP MCP server or be launched directly by the clientDesigned specifically for AI agent use rather than human IDE interaction