Serena
MCP toolkit giving coding agents semantic code retrieval and symbol-level editing via language servers.
About
Serena (oraios/serena on GitHub, ~30K stars) is a powerful open-source MCP toolkit that gives AI coding agents IDE-grade capabilities: semantic retrieval and symbol-level editing powered by language servers (LSP) or an optional JetBrains plugin backend. Instead of reading whole files, agents use tools like get_symbols_overview, find_symbol, and replace_symbol_body to navigate and edit code the way a seasoned developer would — efficiently even in very large projects. It plugs into Claude Code, Codex, OpenCode, Gemini CLI, Cursor, VS Code, JetBrains IDEs, and Claude Desktop via the Model Context Protocol.
Serena also supports project memories (.serena/memories) for persisting findings across sessions. Caveats: the project's license status has been flagged as unclear ("Unlicensed" in some directories) — verify before commercial use; and setup requires uv/uvx plus language-server configuration per project, which adds friction for casual users.
What it can do
- Symbol-level code search and editing via LSP
- MCP server for any compatible AI client
- Project memories across sessions
- JetBrains plugin backend option
- Multi-language semantic analysis
Setup
Install with uvx: uvx --from git+https://github.com/oraios/serena serena start-mcp-server, then register in your MCP client config.
Prerequisites
uv (for uvx); an MCP-capable client (Claude Code, Cursor, etc.); language servers for your languages.
Use cases
Works with
Trust score — 8/10
Scored 8/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 69; Trending score 82/100 from researcher assessment.
★ 30,000 GitHub stars
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.