Serena

31.2K stars
GPL-3.0-or-later (with MIT-licensed SolidLSP component)
Active this month
Last commit 12 days ago

Last verified Nov 9, 2026

Run IDE-level semantic code tools in any AI coding agent via MCP, with symbol-aware retrieval, refactoring, and editing across 40+ languages.

Visit website GitHub31.2K

Read the full breakdownWho it is for, how it works, trade-offs, the quick start and the FAQ

who it's for

Who Serena is for#

Developers running Claude Code or Codex on large multi-file Python projects

Serena gives Claude Code and Codex precise symbol lookup and refactoring tools, replacing token-intensive text searches with single-call operations. Claude Code evaluations on a large Python codebase showed cross-file renames and reference lookups collapsing from 8 to 12 manual steps to one atomic call.

Skip if:

The codebase is small (under a few thousand lines) or the agent is doing mostly generative work, writing new files rather than navigating existing code. Serena's value increases with codebase complexity.

Backend developers maintaining large Java or Kotlin projects in JetBrains IDEs

The JetBrains plugin exposes the full IntelliJ code analysis engine to the agent, including type hierarchy navigation, project dependency search, and interactive debugging with breakpoints. Teams already paying for IntelliJ IDEA can trial the plugin at no initial cost.

Skip if:

You use Rider or CLion; the JetBrains plugin explicitly does not support those IDEs.

Teams building or maintaining multi-language monorepos

With 40+ language server integrations, Serena can provide symbol-level understanding across a monorepo spanning TypeScript, Go, Python, and other languages. Each language server works independently, so you can add language support incrementally.

Skip if:

All relevant code is in one language and your AI client already has adequate built-in code navigation for that language. Adding an MCP server layer adds setup cost without proportional gain in single-language projects.

the problem

The problem it solves#

AI coding agents operate on text. When an agent needs to rename a function, it searches for the string, hopes it catches every reference, and edits each one manually. In a small codebase this works. In a large one, it misses overloaded methods, skips references in generated files, and costs dozens of token-intensive round trips to verify the result.

The challenge deepens with refactoring. Moving a class, inlining a variable, or finding all implementations of an interface requires structural understanding that only a language server or IDE provides. Without it, agents fall back to grep-style searches that return false positives, skip non-obvious usages, and produce edits the developer must then manually verify and fix.

how Serena solves it

How it solves it#

Symbol-Level Code Retrieval

Serena exposes symbol lookup, file outline, reference finding, and declaration navigation as MCP tools. An agent can get all callers of a method as a precise list without scanning every file by text. Both the language server backend and the JetBrains plugin support these operations.

Cross-File Refactoring

Rename and symbol-level editing work via language servers; the JetBrains plugin adds move, inline, and propagate deletions. Unlike text-based search and replace, these operations understand scope, so a rename only touches actual usages of that symbol, not coincidental string matches in comments or strings.

Symbolic Editing Without Line Numbers

Replace a function body, insert code before or after a symbol, or safely delete a symbol with propagated cleanup, all by symbol name. This avoids the off-by-one and context-mismatch errors common to line-number-based editing that most AI agents rely on.

MCP Integration Across Clients

Serena connects to any MCP-compatible client: Claude Code, Codex CLI, OpenCode, Gemini-CLI, VSCode with Copilot, JetBrains IDEs, Claude Desktop, and others. The agent configures a launch command or HTTP URL; no changes to the LLM or client architecture are needed.

40+ Language Support via Language Servers

The default backend uses open-source language servers implementing the LSP protocol, covering Python, TypeScript, Go, Rust, Java, Kotlin, Ruby, PHP, C#, and over 35 other languages. Each language server is independently installed and free to use.

Memory System for Long-Lived Workflows

Serena includes a memory system that persists project context across agent sessions. Notes can be shared across users and projects, reducing warm-up cost on long-running tasks. It complements AGENTS.md-style files and can be disabled if you prefer a different setup.

strengths · trade-offs

Strengths and trade-offs#

Strengths

Works with Any MCP ClientSerena adds semantic capabilities to your existing AI client rather than replacing it. Claude Code, Codex, Cursor, JetBrains AI Assistant, and others all support MCP, so you do not need to switch tools or change your LLM workflow to gain symbol-level code operations.
31,000+ Stars and Active CommunityThe repository has over 31,000 stars and 2,100+ forks as of late 2026, with an active Discord community. Language support across 40+ languages was built in large part by open-source contributors, an unusually wide footprint for an MCP server project.
No Cloud DependencySerena runs entirely locally. Code never leaves your machine during retrieval or refactoring operations. This matters for teams with IP restrictions or compliance requirements that prohibit sending code to third-party services.
GPL-3.0 Permits Commercial and Team UseYou can run Serena in a commercial setting, on internal infrastructure, without licensing fees. The GPL terms require releasing source of modified versions you distribute externally, but running it internally for your own team is unrestricted.

Trade-offs

-Language Server Setup Requires Per-Language InstallationWhile Serena supports 40+ languages, the language servers themselves are separate installs. Some require language-specific toolchains: a working JVM for Java, a Rust toolchain for rust-analyzer, and so on. The Serena documentation lists each dependency, but initial setup on a multi-language project takes more than one uv command.
-Advanced Refactoring Requires the Paid JetBrains PluginMove operations, inline refactoring, type hierarchy navigation, and interactive debugging are only available with the paid JetBrains plugin (free trial available). The free language server backend supports rename and symbolic editing, but the full refactoring surface needs a JetBrains IDE license and the plugin. Rider and CLion are explicitly not supported.
-Active Issue Backlog Across Language IntegrationsThe repository carries 84 open issues. As a tool that integrates with many different language servers and clients, edge cases in language-specific behavior and client compatibility generate a steady volume of reports. Users on less common language stacks may encounter rough edges not present in well-tested languages like Python or TypeScript.

versus alternatives

Serena vs alternatives#

Serena vs GitHub Copilot

GitHub Copilot provides AI-assisted coding via IDE extensions for VS Code, JetBrains, and other editors, including a Workspace feature for codebase-wide search and reference. Copilot is a hosted, subscription-based service at $10 to $19 per user per month.

Feature	Serena	GitHub Copilot
License	GPL-3.0-or-later	Proprietary
Self-hosting	Yes, local MCP server	No, cloud-hosted
Client compatibility	Any MCP-compatible client	Copilot extensions only
Symbol-level refactoring	Yes (rename via LSP, move via JetBrains plugin)	Limited
Interactive debugging	Yes (JetBrains plugin)	No
Language support	40+ via open-source LSP servers	Broad, via extension
Pricing	Free (GPL); JetBrains plugin is paid	$10 to $19 per user per month

Serena is the better fit when you need code operations that work across any MCP-compatible client, not just Copilot extensions, or when code must stay on your machine for compliance reasons. GitHub Copilot is still the better fit for teams who want AI completions and suggestions inline in the editor without configuring a local MCP server, or whose primary workflow is code generation rather than codebase navigation.

Serena vs Cursor

Cursor is an AI-first IDE forked from VS Code. Its codebase indexing embeds the repository and uses semantic search to give the AI agent context from anywhere in the codebase. Cursor is a commercial product with a free tier and paid plans from $20 per month.

Feature	Serena	Cursor
License	GPL-3.0-or-later	Proprietary
Self-hosting	Yes	No
Agent compatibility	Any MCP client	Cursor's own agent
Symbol-level editing	Yes, as user-invokable MCP tools	Built into Cursor's agent, not separately exposed
IDE required	No, works in terminal clients	Yes, the Cursor IDE
Pricing	Free (GPL); JetBrains plugin is paid	Free tier; $20+ per month for higher usage

Serena is the better choice if you do not want to switch IDEs, already use terminal-based agents, or need a self-hosted setup where code stays local. Cursor is the better choice for developers who want an all-in-one AI IDE with semantic search built in and do not need local code operations.

install · quick start

Quick start#

bash

Serena is a Python package requiring Python 3.13. Install with pip or pipx and initialize the project configuration.

pipx install serena-agent
serena init

tech stack · detected from GitHub

What it's built on#

Languages: JavaScriptPython
Frameworks: AngularSvelteVue

frequently asked

FAQ#

Does Serena work with Claude Code?

Yes. Claude Code supports MCP and can launch Serena as a local MCP server via a configured launch command. The Serena documentation covers the specific setup steps for Claude Code. Serena's evaluation suite was tested in part on Claude Code on a large Python codebase, where symbol-level tools reduced multi-step rename operations to single atomic calls.

What is the difference between the language server backend and the JetBrains plugin?

The language server backend uses open-source LSP servers and covers retrieval, diagnostics, rename, and symbolic editing across 40+ languages at no cost beyond the language servers themselves. The JetBrains plugin (paid, free trial available) adds deeper features: move, inline, propagate deletions, type hierarchy, project dependency search, and interactive debugging with breakpoints. The plugin requires IntelliJ IDEA, PyCharm, Android Studio, WebStorm, PhpStorm, RubyMine, or GoLand; Rider and CLion are not supported.

Is Serena free to use commercially?

Yes. The Serena application is licensed under GPL-3.0-or-later, which permits self-hosting and commercial use at no cost. You can run it on your own infrastructure for your team or in a commercial product. If you distribute a modified version externally, you must release those modifications under the same GPL terms. Running it internally or using it commercially without distribution carries no licensing obligation. The JetBrains plugin is a separate paid product with its own pricing.

Can I install Serena without uv?

uv is the only documented prerequisite and the officially supported install path. The README explicitly asks users not to install Serena via MCP or plugin marketplaces because those may contain outdated versions with suboptimal configuration. If you prefer not to use uv, you would need to refer to the source for alternative install paths, which the project does not officially document.

What programming languages does Serena support?

Using the language server backend, Serena supports over 40 languages, including Python, TypeScript, JavaScript, Go, Rust, Java, Kotlin, C#, C/C++, Ruby, PHP, Swift, Scala, Haskell, Lua, and Elixir, among others. The JetBrains plugin supports all languages and frameworks that JetBrains IDEs support, except for Rider (C#) and CLion (C/C++). Some languages require additional toolchain installation beyond uv; the Serena documentation covers these per-language prerequisites.

also worth a look

Similar open-source tools#

Nuphos

Featured

Alternative toConductor

Agent workspace for teams running Claude Code or Codex

372 GitHub stars
Licence Apache-2.0

REA (Reverse Engineer Anything)

Featured

Alternative toRevEng.AI

AI-powered reverse engineering for binaries, apps, and assemblies.

441,247 GitHub stars
Licence MIT

rakazo

Featured

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

7,478 GitHub stars
Licence Apache-2.0

Agent Reach

Featured

Alternative toFirecrawl

Give any AI agent internet access, free and local

109,989 GitHub stars
Licence MIT

Hermes Agent

Featured

Alternative toGrok Bot

Self-hosted AI agent: persistent memory, multi-channel chat, model choice.

261,524 GitHub stars
Licence MIT

Openinterpreter

Alternative toClaude Code

Coding agent for low-cost open models with Codex compatibility

68,842 GitHub stars
Licence Apache-2.0

by category

Open source tools in these categories#

610

12,942 GitHub stars
Licence AGPL-3.0
#Email

rakazo

Alternative toGrok Bot

AI teammates you own: your keys, your model, your machine.

7,478 GitHub stars
Licence Apache-2.0
#AI Agents

SnitchFeed

Live monitoring on LinkedIn, X, Reddit & more. Agentic social listening for startups.

Marketing & Growth

Advertise your product

The projects we add and every new guide. Free, unsubscribe any time.

Find open source alternatives to paid software

Serena vs GitHub Copilot

Feature	Serena	GitHub Copilot
License	GPL-3.0-or-later	Proprietary
Self-hosting	Yes, local MCP server	No, cloud-hosted
Client compatibility	Any MCP-compatible client	Copilot extensions only
Symbol-level refactoring	Yes (rename via LSP, move via JetBrains plugin)	Limited
Interactive debugging	Yes (JetBrains plugin)	No
Language support	40+ via open-source LSP servers	Broad, via extension
Pricing	Free (GPL); JetBrains plugin is paid	$10 to $19 per user per month

Serena vs Cursor

Feature	Serena	Cursor
License	GPL-3.0-or-later	Proprietary
Self-hosting	Yes	No
Agent compatibility	Any MCP client	Cursor's own agent
Symbol-level editing	Yes, as user-invokable MCP tools	Built into Cursor's agent, not separately exposed
IDE required	No, works in terminal clients	Yes, the Cursor IDE
Pricing	Free (GPL); JetBrains plugin is paid	Free tier; $20+ per month for higher usage

Who Serena is for#

Developers running Claude Code or Codex on large multi-file Python projects

Skip if:

Backend developers maintaining large Java or Kotlin projects in JetBrains IDEs

Skip if:

Teams building or maintaining multi-language monorepos

Skip if:

The problem it solves#

How it solves it#

Symbol-Level Code Retrieval

Cross-File Refactoring

Symbolic Editing Without Line Numbers

MCP Integration Across Clients

40+ Language Support via Language Servers

Memory System for Long-Lived Workflows

Strengths and trade-offs#

Strengths

Trade-offs

Serena vs alternatives#

Serena vs GitHub Copilot

Serena vs Cursor

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

Developer Tools

AI & Machine Learning

Find open source alternatives to paid software

Who Serena is for#

Developers running Claude Code or Codex on large multi-file Python projects

Skip if:

Backend developers maintaining large Java or Kotlin projects in JetBrains IDEs

Skip if:

Teams building or maintaining multi-language monorepos

Skip if:

The problem it solves#

How it solves it#

Symbol-Level Code Retrieval

Cross-File Refactoring

Symbolic Editing Without Line Numbers

MCP Integration Across Clients

40+ Language Support via Language Servers

Memory System for Long-Lived Workflows

Strengths and trade-offs#

Strengths

Trade-offs

Serena vs alternatives#

Serena vs GitHub Copilot

Serena vs Cursor

Quick start#

What it's built on#

FAQ#

Similar open-source tools#

Open source tools in these categories#

Developer Tools

AI & Machine Learning

Find open source alternatives to paid software