serena
Serena describes itself as "the IDE for your coding agent". It is an MCP server that gives agents semantic code retrieval, editing, refactoring and debugging tools that operate on symbols rather than line numbers or text search. An agent can find a function's references, rename a symbol across files or move code with one call instead of many fragile text edits, which matters most in large codebases.
It works with any MCP client, including Claude Code, Codex, OpenCode, Gemini CLI, Cursor, VS Code, JetBrains assistants and Claude Desktop. Two backends power the analysis: free language servers (LSP) supporting more than 40 programming languages by default, or a paid Serena JetBrains Plugin that reuses your JetBrains IDE's code analysis. Serena also includes a memory system for long-lived agent workflows.
Getting started: install uv, then run `uv tool install -p 3.13 serena-agent` and `serena init`, and configure the launch command in your client following the official client guide. The project warns against installing it through MCP or plugin marketplaces, whose commands are often outdated.
Features
Why choose it
Trade-offs
Maintenance check
- Last commit
- 2026-10-20
- Commits, last 90 days
- 100+
- Releases, last 90 days
- 3
- Latest release
- v2.1.0 (2026-10-19)
- Open issues
- 122
- Recent issues with at least one comment
- 4/19
Oldest week on the left; the latest four weeks are orange. Source: GitHub commit activity.
Status rule: archived on GitHub = Archived; last commit within 90 days = Active; 91 to 365 days = Slowing; older = Stale. All figures come from the GitHub API on the date shown.
Most discussed open issues
- #1744 zls: rename_symbol silently omits edits in files that are not open (cross-file renames leave stale references)16 comments · 2026-07-23
- #937 Task-1:init_language_server_manager never completes despite JDTLS finishing successfully16 comments · 2026-01-14
- #1491 Discussion: real-usage data on Serena adoption and `find_symbol` fallback patterns (n=21k tool calls, 192 sessions, 21 days)9 comments · 2026-05-15
- #1389 Memory leak in dashboard frontend8 comments · 2026-04-21
- #924 Kotlin LS seems to not provide hover info for a struct without a docstring8 comments · 2026-01-12
Issue titles are quoted from GitHub. Open the link to check whether a problem still applies to the version you use.
Who it fits
- You want to empower LLMs with IDE-level semantic code retrieval, symbol navigation, and surgical multi-file editing capabilities.
- You need an open-source toolkit that turns raw language models into fully capable coding agents operating safely on local repositories.
- You already use commercial AI IDEs like Cursor or Windsurf that have integrated semantic indexing and proprietary diff engines.
- You want a hyper-minimalist 100-line baseline for SWE-bench benchmarking rather than an IDE-like toolset (choose mini-swe-agent).
How it differs
vs codebase-memory-mcp★ 47.7kSerena provides an active toolkit for both semantic retrieval and code editing; codebase-memory-mcp acts as an external vector memory index for search.
vs learn-claude-code★ 79.4kSerena is a production toolkit designed to turn LLMs into coding agents; learn-claude-code is an educational repository for understanding agent concepts.
Compatibility
Quick start
Use cases
Alternatives
Comments
No comments yet. Be the first!