ty-lsp
A Claude Code plugin that integrates
ty (Astral's Python type checker) as an LSP
server, based on the official
pyright-lsp
plugin.
Why?
Without an LSP, Claude Code navigates your codebase by searching text, grepping
for symbol names, scanning files for patterns, and hoping the matches are
relevant. An LSP gives it structured understanding of your code: jump to
definitions, find all references, see type information on hover, and get
real-time diagnostics as it edits. This means Claude can work with your Python
code the way an IDE does, following the actual structure of the code rather than
guessing from string matches.
The pyright-lsp plugin does all of this too! So why ty? There a few reasons,
one is that ty is written in Rust with incremental recomputation at its core,
so in practice it uses less CPU and memory than Pyright, and tends to be faster
at updating diagnostics after edits. See the
ty benchmarks for
detailed comparisons.
In practice, Claude Code's response time is dominated by LLM reasoning calls, so
the raw speed difference between type checkers matters less than in other
contexts. What matters more is that the LSP your agent is using is consistent
with the type checker you run in CI. If you use ty in CI, this plugin means
Claude sees the same view of your code as your pipeline does.
Prerequisites
Install ty globally so that it is available on your PATH. The recommended
approach is with uv:
uv tool install ty
Any other method of globally installing ty should also work, such as
pipx, the
Astral install script,
or even a global pip install ty (though that's not a great idea).
Installation
Install from the
tcbegley-cc-plugins
marketplace:
/plugin marketplace add tcbegley/tcbegley-cc-plugins
/plugin install ty-lsp@tcbegley-cc-plugins
Development
claude --plugin-dir /path/to/ty-cc-plugin
What it provides
- Real-time Python type checking diagnostics as Claude edits code
- Code navigation (go to definition, find references)
- Hover information with type details
Configuration
ty is configured via ty.toml or the [tool.ty] section of pyproject.toml.
See the ty configuration docs for
details.
K-dash
boostvolt
TheBushidoCollective