AI Tools

Flagify ships with first-class integrations for the AI coding tools you already use. Manage feature flags, create targeting rules, and toggle environments without leaving your AI workflow.

Supported tools

Quick setup

Every integration follows the same pattern: drop a config file in your repo, and the AI tool picks it up automatically.

# Install the Flagify CLI
npm install -g @flagify/cli

# Generate AI tool configs for your project
npx flagify ai-setup

This creates the appropriate config files based on the AI tools detected in your environment.

Claude Code

Slash commands

Add custom slash commands to your .claude/commands/ directory:

Create a flag:

<!-- .claude/commands/flagify-create.md -->
Create a new feature flag called "$FLAG_NAME" with type
"$FLAG_TYPE" (boolean, string, number, json).

Use the Flagify CLI:
flagify flags create $FLAG_NAME --type $FLAG_TYPE --description "$DESCRIPTION"

Toggle a flag:

<!-- .claude/commands/flagify-toggle.md -->
Toggle the flag "$FLAG_NAME" in the "$ENVIRONMENT" environment.

Use the Flagify CLI:
flagify flags toggle $FLAG_NAME --environment $ENVIRONMENT

List all flags:

<!-- .claude/commands/flagify-list.md -->
List all feature flags in the current project.

Use the Flagify CLI:
flagify flags list

CLAUDE.md context

Add Flagify context to your CLAUDE.md so the agent understands your flag setup:

<!-- CLAUDE.md -->
## Feature Flags

This project uses Flagify for feature flag management.

- SDK: `@flagify/node` (server) and `@flagify/react` (client)
- CLI: `flagify flags list`, `flagify flags create <key>`, `flagify flags toggle <key>`
- Environments: development, staging, production
- Flag evaluation is local (cached) with streaming sync
- Use `isEnabled()` for boolean flags, `getValue()` for typed values
- Flag names use kebab-case: `new-checkout-flow`

Cursor

.cursorrules

Create a .cursorrules file in your project root:

# Flagify Feature Flags

This project uses Flagify (@flagify/node, @flagify/react) for feature management.

## SDK Patterns

When creating flag checks:
- Import from '@flagify/node' for server code, '@flagify/react' for client
- Use isEnabled() for boolean flags
- Use getValue<T>() for typed values (string, number, JSON)
- In React, use useFlag(), useVariant(), useFlagValue() hooks

## Code Patterns

Server-side flag evaluation:
const isEnabled = flagify.isEnabled('flag-name');

React component with flag:
import { useFlag } from '@flagify/react';
const showFeature = useFlag('flag-name');

## Rules
- Never hardcode flag values — always evaluate dynamically
- Flag names use kebab-case: 'new-checkout-flow'
- Group related flags with a shared prefix: 'checkout-*'
- Always handle both flag states (enabled and disabled)

GitHub Copilot

copilot-instructions.md

Create .github/copilot-instructions.md:

# Flagify Feature Flags

This project uses the Flagify SDK for feature flag management.

## When writing code that involves feature flags:

1. Import from @flagify/node (server) or @flagify/react (client)
2. Use isEnabled(flagKey) for boolean flags
3. Use getValue<T>(flagKey) for typed values
4. In React: useFlag(), useVariant(), useFlagValue() hooks
5. Always handle both enabled and disabled states
6. Flag names are kebab-case: 'feature-name'

## Environments: development, staging, production

Prompt files

For more complex workflows, use .github/copilot/ prompt files:

<!-- .github/copilot/create-flag-check.md -->
Create a feature flag check for the given feature.
Use the Flagify SDK with proper error handling.
Include both the enabled and disabled code paths.
Use the existing flagify client from 'src/lib/flagify'.

Windsurf

.windsurfrules

Create a .windsurfrules file in your project root:

# Flagify Feature Flags

Project uses Flagify for feature management.
SDK: @flagify/node (server), @flagify/react (client)

## Patterns
- Boolean flags: flagify.isEnabled('flag-name')
- Typed values: flagify.getValue<string>('flag-name')
- React: const enabled = useFlag('flag-name')

## Conventions
- Flag names: kebab-case ('new-feature')
- Always handle both states
- To check current flags: flagify flags list

Best practices

Flag naming in AI contexts

Use descriptive, kebab-case names that help the AI understand intent:

// Good — AI understands the purpose
'new-checkout-flow'
'premium-dashboard-v2'
'enable-dark-mode'

// Avoid — ambiguous for AI
'flag1'
'test'
'temp'

Context files

Keep a FLAGIFY.md or include flag documentation in your README.md so AI tools understand your flag setup:

## Active flags

| Flag | Type | Purpose | Owner |
|------|------|---------|-------|
| `new-checkout-flow` | boolean | Checkout redesign rollout | @team-payments |
| `pricing-page-v2` | string | A/B test pricing layout | @team-growth |

Generating setup automatically

The Flagify CLI can scaffold all AI tool configs at once:

# Generate all configs
npx flagify ai-setup

# Generate for a specific tool
npx flagify ai-setup --tool claude

# Include a snapshot of current flags (AI tools always get a
# "run flagify flags list" instruction for live data too)
npx flagify ai-setup --include-flags