Open sourcePython · ≈31k stars · GPL-3.0-or-later (SolidLSP: MIT)
Serena
MCP toolkit giving coding agents IDE-like semantic retrieval, editing and refactoring.
Serena is an MCP toolkit that gives coding agents semantic code retrieval, editing, refactoring and debugging tools similar to an IDE's. The tools work at the symbol level and use the code's relational structure instead of line numbers or plain text search. It integrates with any MCP client or LLM. Serena v2 added a REPL-based, code-execution agent interface. The project publishes agent-written evaluations from Claude Code and Codex CLI runs on large codebases. Licensing is per component: the SolidLSP language-server layer is MIT and the Serena application is GPL-3.0-or-later.
Pros & cons
Each point cites its sources and says what kind of evidence backs it. A second reviewer tried to refute every claim before it was kept.
Pros
- DocsGives any MCP client symbol-level retrieval, editing and refactoring through LSP for 40+ languages, instead of line-number text edits.[1]
- User reportsAn optional JetBrains plugin backend uses the IDE's own code analysis, and one user says it exposes more than the standard IDEA MCP.[1][2]
- User reportsUsers pair it with OpenCode and with cheaper models (Gemini Flash) to make up for weak built-in tools.[1][2]
- MetadataActively developed: v2.1.0 (2026-10-19) added a REPL-based agent interface.[1]
Cons
- MetadataThe Serena application is GPL-3.0-or-later (only SolidLSP is MIT), so embedding or redistributing it carries copyleft obligations.[1]
- User reportsIn 2025, users reported that replace_symbol_body often duplicated code and that connecting it to Claude Code failed (41- and 43-comment issues, both since closed).[1][2]
- DocsThe JetBrains backend is a paid plugin (free trial) and does not support Rider or CLion.[1]
- DocsIts headline evaluation has agents rate Serena's own tools, not an independent benchmark.[1]
- Best for
- Agents working in large, multi-language codebases where symbol-aware navigation and cross-file refactors beat grep-and-patch.
- Watch out for
- Check the GPL terms before bundling it, and watch symbol-replace edits for duplicated code.
Assessed Oct 29, 2026 · adversarially reviewedSuggest a correction
In research (5)
- Stratum: Repository Memory for Coding Agents Beyond the Context WindowVerdict: Incremental
Oct 26, 2026 · Semantic code retrieval tools for coding agents via MCP, an alternative approach to the selection problem.
- RepoTrace: Improving Requirement Tracing in Multi-Round Agent InteractionsVerdict: Incremental
Oct 26, 2026 · Symbol-level code navigation for agents, another way to focus edits.
Oct 26, 2026 · Symbol-level retrieval toolkit addressing the same 'find the right helpers and callers' problem.
- How Do Small Coding Agents Use Code Graphs and Embeddings?Verdict: Important
Oct 25, 2026 · Semantic code-navigation tools for agents; the findings on tool uptake and name lookup apply directly.
- Code Comprehensor: Hierarchical Documentation and Cross-Modal Retrieval for Large RepositoriesVerdict: Incremental
Oct 21, 2026 · MCP toolkit for semantic code retrieval, a competing approach.
Neighbours on the workbench
TypeScript · ≈393k stars · MIT
openclaw/openclaw
The viral open-source personal AI assistant that really does things.
Maintained by OpenClaw Foundation
MDX · ≈307k stars · MIT
obra/superpowers
An agentic skills framework and software methodology that works.
Maintained by Jesse Vincent (obra)
Shell · ≈306k stars · MIT
mattpocock/skills
Small, composable agent skills for real engineering work.
Maintained by Matt Pocock (AI Hero)
Connections
Relations from the Harness Hub knowledge graph, as charted in the State of Harness reports.
Maintained by · 1
- Oraios AIOrganisation
Built on · 3
- Claude CodeProject
- CodexProject
- Model Context ProtocolProject
Similar entries · 1
- codebase-memory-mcpProject
Sources
Verified Oct 29, 2026