AI-powered code pattern extraction, conformance checking, and deep analysis for VS Code.
Codelab analyzes your codebase to extract engineering patterns, then enforces them — catching style violations, implementation issues, and deeper problems before they reach production. It uses Claude Code CLI as its AI backbone.
Built for teams who care about code quality but don't want to write 500 ESLint rules by hand.
Every codebase has patterns — naming conventions, error handling styles, import ordering, architectural decisions. New team members break them. AI coding assistants ignore them. Linters only catch what you explicitly configure.
Codelab solves this by:
- Extracting your existing patterns automatically from the code itself
- Checking every file against those patterns
- Showing violations inline with squiggly lines, hover cards, and quick-fix suggestions
- Blocking sloppy code from being pushed via a git pre-push hook
- Finding deeper issues like redundancy, race conditions, performance bottlenecks, and security concerns
No manual rule writing. No YAML configuration. Just point it at your code and go.
Codelab requires Claude Code CLI to be installed and authenticated. Without it, the extension cannot function.
Install Claude Code:
npm install -g @anthropic-ai/claude-codeThen authenticate:
claudeFollow the prompts to sign in with your Anthropic account.
If Claude Code CLI is not detected, Codelab will show an error with a link to the install guide.
| Feature | Minimum Plan |
|---|---|
| Pattern Extraction | Any Claude plan |
| Conformance Checking | Any Claude plan |
| Push Guard | No API calls (local) |
| Deep Analysis | Pro or higher (uses Opus model) |
Deep Analysis uses the Opus model for orchestration, which requires a Claude Pro subscription or higher. A cost warning is shown before running.
Install Codelab from the VS Code Marketplace, then reload your window.
Open the Command Palette (Cmd+Shift+P / Ctrl+Shift+P) and run:
Codelab: Extract Patterns
This analyzes your codebase and generates:
.codelab/patterns.md— your codebase's engineering patterns.codelab/CLAUDE.md— rules for Claude Code to follow.codelab/.cursorrules— rules for Cursor AI to follow
For large codebases (50+ files), Codelab automatically deploys parallel sub-agents to analyze chunks concurrently, then compiles the results into a unified document.
Run:
Codelab: Check Code Conformance
This checks every file against your extracted patterns and reports violations as:
- Inline squiggly lines (errors and warnings)
- Hover cards with the rule and suggested fix
- Quick-fix lightbulb actions
- A sidebar tree view grouped by file
As you fix violations and save files, Codelab automatically removes resolved issues in real-time — no need to re-run the full check.
Run Codelab: Toggle Push Guard to install a git pre-push hook. When enabled, git push will be blocked if there are conformance errors (warnings are allowed through).
Run Codelab: Run Deep Analysis for a thorough review that finds:
- Redundancy — duplicate logic, copy-pasted code
- Implementation issues — race conditions, incorrect async patterns, resource leaks
- Performance — N+1 queries, blocking operations, expensive hot paths
- Security — injection risks, hardcoded secrets, insecure defaults
Results appear in the Deep Analysis panel in the Codelab sidebar, grouped by severity (critical / warning / info).
Automatically discovers your codebase's naming conventions, import styles, error handling patterns, formatting, type usage, and more. Outputs a structured patterns.md document.
Checks all files against extracted patterns. Reports issues with file, line, severity, and a specific suggestion for how to fix each one.
When you save a file, Codelab checks if flagged lines have changed and removes resolved issues instantly — without calling Claude again.
Conformance results survive window reloads. Issues are cached in .codelab/conformance.json with SHA-256 hash validation. If files or patterns change, stale issues are automatically discarded.
A pre-push hook that reads the cached conformance data and blocks git push if errors exist. No API calls at push time — it's instant.
For codebases with 50+ files, Codelab chunks files by directory and spawns parallel Haiku sub-agents. A Sonnet compiler then merges the results. Chunk files are saved to .codelab/patterns/ for audit.
An Opus-orchestrated, multi-agent analysis that runs 4 specialized Sonnet agents in parallel (redundancy, implementation, performance, security). Results are compiled and shown in a dedicated tree view.
All long-running operations (pattern extraction, conformance checking, deep analysis) can be cancelled from the progress notification.
| Command | Description |
|---|---|
Codelab: Extract Patterns |
Analyze codebase and generate patterns.md |
Codelab: Check Code Conformance |
Check all files against patterns |
Codelab: Run Deep Analysis |
Run Opus-orchestrated deep analysis |
Codelab: Toggle Push Guard |
Enable/disable git pre-push hook |
Codelab: Count Workspace Files |
Count analyzable files |
All generated files live in the .codelab/ directory:
.codelab/
patterns.md — extracted engineering patterns
CLAUDE.md — rules for Claude Code
.cursorrules — rules for Cursor AI
conformance.json — cached conformance issues
settings.json — extension settings
patterns/ — sub-agent chunk outputs (large codebases)
chunk-001.md
chunk-002.md
...
analysis/ — deep analysis outputs
redundancy.md
implementation.md
performance.md
security.md
summary.md
You can choose to .gitignore the .codelab/ directory or commit it so your team shares the same patterns.
"Claude Code CLI not found"
Install it with npm install -g @anthropic-ai/claude-code and make sure claude is in your PATH.
"No patterns found"
Run Codelab: Extract Patterns before checking conformance.
Deep Analysis button is grayed out or fails Deep Analysis requires a Claude Pro or higher subscription for Opus model access.
Conformance results disappeared after reload If you modified files or re-extracted patterns, cached issues are invalidated. Run conformance check again.
Initial release:
- Pattern extraction with auto-scaling (single session or parallel sub-agents)
- Conformance checking with inline diagnostics, hover cards, and code actions
- Real-time issue resolution on file save
- Persistent conformance cache with hash validation
- Git pre-push guard (toggle on/off)
- Deep Analysis engine (Opus + Sonnet multi-agent)
- Deep Analysis tree view grouped by severity
- Cancellable operations
- Claude Code CLI detection
Brian Mwangi — brianmwangi.dev
MIT