
Ctxlint
Catch stale or wrong instructions in CLAUDE.md and AGENTS.md before they mislead your coding agent on every session.
Overview
io.github.YawLabs/ctxlint is a Ship-phase MCP server that lints agent context files against your actual codebase.
What is this MCP server?
- Lints CLAUDE.md, AGENTS.md, and related agent context files against the codebase
- @yawlabs/ctxlint v0.14.1 as stdio MCP server
- Surfaces drift between documented conventions and actual project layout
- Supports multi-phase use: docs during build, review before ship, cleanup in operate
- npm-installable for Claude Code and Cursor MCP configs
- Published server version 0.14.1
- npm package @yawlabs/ctxlint
- Targets CLAUDE.md, AGENTS.md, and related agent context files
What problem does it solve?
Agent instruction files go stale after refactors, so the model follows wrong paths and rules on every task.
Who is it for?
Builders who maintain agent context markdown and want a fast drift check before trusting the agent on critical work.
Skip if: Projects with no agent context files or teams that never customize CLAUDE.md-style guidance.
What do I get? / Deliverables
You get concrete lint feedback aligning CLAUDE.md and AGENTS.md with what the repo actually contains.
- Lint report comparing context file claims to repository reality
- Actionable fixes for stale paths, commands, and conventions
- Higher-trust agent sessions after context alignment
Recommended MCP Servers
Journey fit
How it compares
Context-file linter MCP, not a general ESLint replacement or cloud deploy integration.
Common Questions / FAQ
Who is io.github.YawLabs/ctxlint for?
Solo developers using Claude Code, Cursor, or similar agents who keep CLAUDE.md or AGENTS.md in sync with a living codebase.
When should I use io.github.YawLabs/ctxlint?
Run it before ship review, after large refactors, or whenever agent behavior seems to ignore your documented project rules.
How do I add io.github.YawLabs/ctxlint to my agent?
Install @yawlabs/ctxlint from npm, add it as a stdio MCP server in your host config, and invoke lint tools against your repo root where context files live.