Build clearer agent instructions #

LintLang checks the files that define how your agent behaves: instructions, prompts, tool descriptions, and schemas. Get actionable feedback before those changes reach a model.

Start hereScan your project →Discover the agent instructions in your repository and understand the results. Your workflowChoose an integration →Use LintLang in your editor, agent host, local hooks, or CI.

One command to get started #

From your project root, with uv and Python 3.10+:

uvx --from lintlang==0.8.0 lintlang scan --discover .

The scan runs locally without model calls. It finds recognized instruction files across your project; add other configuration files explicitly. Follow the quickstart for the complete workflow, or choose a persistent installation.

Make the next change easier #

Understand a finding. The rule explorer has examples and repairs for each diagnostic.
Control what gets checked. Learn about inputs and discovery, then baseline existing findings as you adopt LintLang.
Review changes automatically. Choose a report format and CI threshold, or use a ready-made integration.
Try a small repair. Work through the before-and-after example.

Reference and project #

Find every option in the CLI reference, and the detailed behavior in the full reference. Coverage and limitations explains how to interpret a clean result.

Release notes · Updates and community · Contribute · Security.