MCP et appels d’outils

Serena: A Semantic Code Retrieval and Editing MCP Toolkit for AI Agents

oraios/serena

Serena is an MIT-licensed, Python-based MCP toolkit that provides AI coding agents with IDE-grade semantic code retrieval, editing, refactoring, and debugging capabilities through high-level symbol-level abstractions.

★ 27,6KÉtoiles
⑂ 1,8KForks
107Problèmes ouverts
PythonLangue
MITLicence
Q@project.QualityScoreScore éditorial

Captures d'écran du projet

Capture d’écran de Serena: A Semantic Code Retrieval and Editing MCP Toolkit for AI Agents Capture d’écran de Serena: A Semantic Code Retrieval and Editing MCP Toolkit for AI Agents

Vue d’ensemble

Serena acts as an IDE layer for coding agents, offering semantic code retrieval, editing, refactoring, and debugging tools that operate at the symbol level rather than relying on primitive line-number or text-search operations. It integrates with AI clients via the Model Context Protocol (MCP) and supports two backends: open-source language servers (LSP) supporting over 40 programming languages by default, and a paid JetBrains plugin that leverages JetBrains IDE analysis capabilities. Serena's agent-first design aims to make agents faster, more efficient, and more reliable, particularly on large and complex codebases.

Fonctionnalités clés

  • Semantic code retrieval: find symbol, file outline, find referencing symbols, find declaration, find implementations, diagnostics
  • Refactoring: rename, move, inline, propagate deletions (JetBrains plugin only for move, inline, and propagate deletions)
  • Symbolic editing: replace symbol body, insert after/before symbol, safe delete
  • Interactive debugging with breakpoints and REPL-style execution control (JetBrains plugin only)
  • Memory management system for persistent, cross-session agent workflows
  • Multi-layered YAML configuration: global, CLI, per-project, execution context, and composable modes
  • Support for 40+ programming languages via LSP language servers
  • Basic utilities: regex search, text replacement, directory listing, file search, file reading, shell command execution
  • MCP server with launch command or HTTP mode connectivity

Prérequis, installation et démarrage rapide

Install Serena using uv with the command: `uv tool install -p 3.13 serena-agent`. After installation, the `serena` command should be available. The README explicitly warns against installing Serena via an MCP or plugin marketplace, as they contain outdated and suboptimal installation commands.

Utilisation

After installation, run `serena init` to initialise Serena and verify the setup. By default, this configures the language server backend; to use the JetBrains backend instead, add the parameter `-b JetBrains`. Then, configure your MCP client by providing it with a Serena launch command or an HTTP server URL. Specific client configuration instructions are available in the documentation.

Compatibilité des modèles et cas d’usage

Serena is client and LLM agnostic, integrating via MCP. It is compatible with terminal-based clients (Claude Code, Codex, OpenCode, Gemini CLI), IDE plugins (Copilot, Junie, JetBrains AI Assistant for VSCode, Cursor, JetBrains IDEs), and desktop/web clients (Claude Desktop, Codex App, OpenWebUI).

Licence et notes sur les risques

Licensed under the MIT License.

Editorial verification 2026-08-09: repository URL, owner, description, license and repository statistics were reviewed. License metadata: MIT. README was fetched for the channel draft; re-check repository dependencies, releases and model terms before production use.

Publication et maintenance

Repository created on 2025-03-23. Last pushed on 2026-08-05. Default branch is main. The repository is not archived.

MemPalace

mempalace/mempalace

★ 58KPython

QwenPaw

agentscope-ai/qwenpaw

★ 33,9KPython

Composio

composiohq/composio

★ 29,5KTypeScript