codegraph vs serena
Two of the top projects in MCP servers for developers, side by side: who leads on each score part and who each one suits, then setup, license, activity and what each review found.
codegraph
Local code knowledge graph served to coding agents over MCP
serena
MCP server giving coding agents symbol-level code search, editing and refactoring
Verdict
codegraph ranks higher, 81 to 71. Of the 5 score parts we measure for both, codegraph leads on 3, serena on 1 and 1 is even.
- Adoptioncodegraph leads by 25 points, 95 to 70: a wide gap.
- FreshnessEven at 100.
- Maintenanceserena leads by 6 points, 95 to 89.
- Easy to runcodegraph leads by 17 points, 50 to 33: a wide gap.
- Agent-readycodegraph leads by 15 points, 85 to 70: a wide gap.
Pick codegraph if you are
Developers using AI coding agents on mid-to-large codebases
- Runs locally; no Node.js needed, ships as a bundled installer or npm package
- One `codegraph install` configures MCP for nine named agents and IDEs
Pick serena if you are
Developers who want their coding agent to navigate and refactor large codebases
- Symbol-level find, reference lookup and edit tools instead of line-number or regex operations
- Language server backend covers 40+ languages and is free to use
| What we compare | codegraph | serena |
|---|---|---|
| Score parts, out of 100 | ||
| Adoption | 95, widely used | 70, popular |
| Freshness | 100, active | 100, active |
| Maintenance | 89, healthy | 95, healthy |
| Easy to run | 50, easy | 33, some setup |
| Agent-ready | 85, ready | 70, partly |
| Facts from GitHub and the README | ||
| Stars | 74.1k | 30.3k |
| License | MIT (permissive) | custom license (read the license) |
| Last commit | Oct 2026 | Oct 2026 |
| Last release | Oct 2026 | Oct 2026 |
| Language | C | Python |
| Docker | No | Yes |
| GPU | Not needed | Not needed |
| arm64 or Apple Silicon | Mentioned | Not stated |
codegraph
CodeGraph indexes a project into a local graph of symbols, call edges and dependencies under `.codegraph/`, then exposes it to coding agents through an MCP server. Agents can fetch relevant source, call paths and change impact in one call instead of grepping and reading files. `codegraph install` wires it into Claude Code, Cursor, Codex CLI, opencode, Gemini CLI, Kiro, GitHub Copilot and others, and auto-sync keeps the index current.
Who it is for: Developers using AI coding agents on mid-to-large codebases
Strengths
- Runs locally; no Node.js needed, ships as a bundled installer or npm package
- One `codegraph install` configures MCP for nine named agents and IDEs
- Covers 30+ languages, including TypeScript, Python, Go, Rust, Java, C#, Swift
- Watches files and updates the graph automatically; no manual re-index
Weaknesses
- README admits about 80% more retrieval context stays resident in long sessions
- Published benchmarks are the author's, on 7 repos with one question each
- No Docker or compose setup; install is via curl/PowerShell script or npm
- Hosted CodeGraph platform is a waitlist product, separate from this repo
- no GPU
serena
Serena is an MCP server that exposes IDE-style tools to coding agents: find symbol, find referencing symbols, symbol-body replacement, rename, and a memory system. It uses language servers (LSP, 40+ languages) by default, or a paid JetBrains plugin that adds move, inline, type hierarchy and debugging. It works with Claude Code, Codex, Cursor, VSCode, Claude Desktop and other MCP clients.
Who it is for: Developers who want their coding agent to navigate and refactor large codebases
Strengths
- Symbol-level find, reference lookup and edit tools instead of line-number or regex operations
- Language server backend covers 40+ languages and is free to use
- Works with any MCP client, including Claude Code, Codex, Cursor and OpenWebUI
- Layered YAML configuration: global, per-project, per-client context and modes
Weaknesses
- Move, inline, type hierarchy and debugging need the paid JetBrains plugin
- LSP backend rename covers symbols only; some languages need extra dependencies installed
- Mixed licensing: GPL-3.0-or-later application with MIT SolidLSP component
- README warns against marketplace installs; setup requires uv and per-client configuration
- no GPU
- Docker + Compose
- Needs uv, language servers (per language), JetBrains IDE plugin (optional, paid)
- Models: any MCP-capable client/LLM
More in MCP servers for developers
Building with these?
We help teams pick an open-source AI stack and run it in production. Client work never changes a rank.