# Serena — Semantic Code Agent via MCP

> Serena is a coding agent MCP server with semantic code retrieval and editing. 22K+ stars. Understands classes, functions, symbols for precise context-aware edits. MIT.

## Install

Merge the JSON below into your `.mcp.json`:

# Serena — Semantic Code Agent via MCP

> A coding agent MCP server that understands code structure — classes, functions, symbols — for precise, context-aware edits. 22K+ GitHub stars.

## Quick Use

```bash
pip install serena
serena init
```

Add to your Claude Code MCP config (`.claude/mcp.json`):
```json
{
  "mcpServers": {
    "serena": {
      "command": "serena",
      "args": ["serve", "--project", "."]
    }
  }
}
```

Now Claude Code can navigate, understand, and edit your codebase at the symbol level — not just files.

## Intro

Serena is a semantic code retrieval and editing toolkit that works as an MCP server. Unlike basic filesystem MCP tools that operate on raw files, Serena understands your code's structure:

- **Symbol-level navigation**: Jump to class definitions, function signatures, method implementations across files
- **Semantic search**: Find code by meaning, not just string matching — "find the authentication middleware" works
- **Precise edits**: Edit specific functions or classes without touching surrounding code
- **Multi-language**: Python, TypeScript, JavaScript, Go, Rust, Java, and more via tree-sitter
- **Context windows**: Automatically provides relevant context (imports, types, related functions) to the AI agent

This makes AI coding agents significantly more effective — instead of reading entire files, they can surgically navigate to exactly what they need.

## FAQ

**Q: How is this different from the Filesystem MCP Server?**
A: Filesystem MCP operates on raw files (read/write/list). Serena operates on code symbols (classes, functions, imports). It's the difference between a text editor and an IDE.

**Q: Does it work with Claude Code?**
A: Yes, as an MCP server. Add it to your `.claude/mcp.json` and Claude Code gains semantic code understanding.

**Q: What languages does it support?**
A: Python, TypeScript/JavaScript, Go, Rust, Java, C/C++, Ruby, and more. Uses tree-sitter for parsing, so language support is broad and accurate.

**Q: Does it require an index?**
A: It builds a lightweight index on first run (`serena init`). Incremental updates are fast — typically < 1 second for file changes.

## Works With

- Claude Code (MCP server)
- Cursor, Windsurf (MCP integration)
- Any MCP-compatible AI agent
- Python, TypeScript, Go, Rust, Java, C/C++, Ruby

## Source & Thanks

- GitHub: https://github.com/oraios/serena (22K+ stars)
- License: MIT
- Docs: https://serena.dev


---
Source: https://tokrepo.com/en/workflows/5c63d696-1d73-4e33-994a-8d7d7151c640
Author: MCP Hub