Serena
MCP coding toolkit that runs locally and connects to Claude Code, Codex and other clients, with language-server tools and a JetBrains backend.
Serena is a locally run MCP toolkit that gives AI coding agents access to code structure: functions, classes and the references between them. It's for developers who want their agent to find and change specific parts of a codebase without reading whole files or relying on text matching. It requires an LLM client and works with Claude Code, Codex, OpenCode, Cursor and OpenWebUI.
Agents can look up symbols, inspect file outlines, follow references and check diagnostics. Editing tools replace a function or class body, insert code beside a symbol and rename symbols across references. This focus on code relationships makes Serena relevant to large projects where a change can affect several files. A memory system retains project knowledge for use across sessions, users and projects.
The language-server backend supports languages including Python, TypeScript, Java, Rust and Swift. Some capabilities, such as finding implementations, depend on the language server. The JetBrains plugin uses the analysis tools in IDEs such as IntelliJ IDEA and PyCharm, adding dependency searches, type hierarchies and refactorings for moves and inlining. It also lets agents debug with breakpoints, inspect variables and control execution. Rider and CLion aren't supported.
Serena supplies the tools; the connected LLM directs their use. Running its server locally doesn't make a cloud-connected client offline. The open-source application uses GPL-3.0-or-later, while its SolidLSP component uses MIT; combined distributions fall under the GPL.
