Open source · ISC license · No telemetry

Your AI context files are lying to you.

ez-context detects when CLAUDE.md, Cursor rules, and Copilot instructions drift from what your code actually does — then fixes them.

npm install -g @ez-corp/ez-context

See how it works

Context rot is real. And it's probably happening to you right now.

Your context files drift. Every refactor, every rename, every architectural decision widens the gap between what your AI reads and what your code does. You started with Express.js. You migrated to Fastify six months ago. Your CLAUDE.md still says Express.

CLAUDE.md says

## Tech Stack

- Language: TypeScript

- Framework: Express.js stale

- REST API patterns

Your code does

import Fastify from 'fastify';

import { fp } from 'fastify-plugin';

const app = Fastify ({ logger: true });

Your AI assistant is now suggesting patterns that don't exist in your codebase. Hallucination-adjacent, but entirely your fault. ez-context finds the gaps and patches them.

Watch it work.

Two commands. One for generation, one for drift detection.

ez-context

$ ez-context generate

Analyzing project structure...

Extracting conventions from 47 files...

Generated CLAUDE.md (2.1kb)

Generated AGENTS.md (1.8kb)

$ ez-context drift

Checking 12 claims against codebase...

Health score: 94/100

2 claims need attention

Everything your context files need.

🔍

Smart Extraction

Analyzes package.json, tsconfig, CI configs, and code patterns. No manual config.

📡

Drift Detection

Compares every claim in your context files against actual code using semantic search.

✏️

Targeted Updates

Rewrites only drifted sections. Your manual notes stay untouched.

📄

7 Output Formats

CLAUDE.md, AGENTS.md, Cursor rules, Copilot instructions, SKILL.md, Rulesync, Ruler.

🌐

Multi-Language

TypeScript, Go, and Rust. More coming.

📊

Confidence Scores

Every extracted convention includes a confidence score and evidence references.

Drift happens. We catch it.

What your context file says

## Tech Stack

✓ Language: TypeScript

Framework: Express.js STALE

REST API patterns STALE

Middleware-based routing STALE

What your code does

import Fastify from 'fastify';

import { fastifyPlugin } from

'fastify-plugin';

const app = Fastify({

logger: true

});

Context health score 0/100

After running ez-context drift

One tool. Every format.

Generate all 7 formats at once, or pick exactly what you need.

CLAUDE.md

Anthropic Claude Code context file. Used by claude-code and MCP-compatible tools.

AGENTS.md

Linux Foundation agent standard. Cross-platform context for any agent runtime.

Cursor Rules

.cursor/rules/*.mdc

Cursor IDE rules in MDC format with glob-scoped context per rule.

Copilot

.github/copilot-instructions.md

GitHub Copilot workspace instructions. Injected into every Copilot interaction.

SKILL.md

Skill modules with progressive disclosure. Structured for agent skill registries.

Rulesync

.rulesync/*.md

Rulesync format for cross-editor rule synchronization and sharing.

Ruler

.ruler/*.md

Ruler format for opinionated context organization with tagging support.

Get started in 10 seconds.

Free. Open source. ISC license. No accounts required.

npm install -g @ez-corp/ez-context

ez-context generate Generate context files from your codebase

ez-context drift Check your existing files for drift

GitHub npm package

Free. Open source. ISC license. No accounts, no telemetry.

Frequently asked questions.

Is it free?

Yes. Free, open source, ISC license. No accounts, no telemetry, no BS.

Does it work with my editor?

If your editor reads CLAUDE.md, .cursorrules, or copilot-instructions.md, yes. We generate for Claude Code, Cursor, GitHub Copilot, and more.

What languages does it support?

TypeScript, Go, and Rust. More languages are coming. The static analysis works on any project with a package.json or go.mod.

Does it phone home?

No. Everything runs locally. Your code never leaves your machine.

How does drift detection work?

We parse claims from your context files, embed them, and compare against your codebase using semantic search. Each claim gets a confidence score.