Code Reference

If you want to use Serena — connect it to your coding agent, configure languages, work with projects — this is not the page you need. Start with Getting Started and come back only if you end up reading the source. Nothing here is required for using Serena.

Serena is a server, not a library, and this is not an API reference: nothing below is a published interface, and any of it may change between releases without notice. What Serena does offer as an interface is its set of tools, which is documented in Tools.

This section is for reading and extending the code: your first contribution, embedding Serena in your own agent, or understanding what a tool actually does under the hood. It is generated from the docstrings in src/serena, and every entry links to its source — treat it as a guided way of reading the code, not a substitute for it.

The shape of the package, in the order you’ll meet it

1. The agent — serena.agent holds SerenaAgent: it loads the configuration, activates a project, starts language servers, and decides which tools exist. Everything else hangs off it.

2. The tools — every capability an LLM sees is a Tool subclass. The base machinery (execution, logging, error recovery) lives in serena.tools.tools_base; the symbolic operations you’d actually recognize — find_symbol, replace_symbol_body — are in serena.tools.symbol_tools. A tool’s docstring is its prompt: what you read there is exactly what the LLM reads.

3. The configuration — serena.config.serena_config contains the dataclasses behind serena_config.yml and project configuration.

4. The project layer — serena.project is file access, ignore handling, and the bridge to the language-server library.

Want to build your own agent on top of Serena? The custom agent guide is the narrative version of this section.

Undocumented members are deliberately left out — every entry you find here carries an actual explanation. (The solidlsp language-server library is not part of the generated reference — SolidLSP is its guided tour.)