Deploy feature flags in three steps: install the SDK, initialize the client, and evaluate your first flag. FlagPilot evaluates flags from an in-process local cache — no database round-trips at request time.
Prerequisites
- A FlagPilot account and API credentials
- Node.js 18+ for JavaScript SDKs, Python 3.8+ for Python, or Ruby 3.0+ for Ruby
- A project on your FlagPilot dashboard
Install the SDK
Install the @flagpilot/node package from npm:
npm install @flagpilot/node
Quick start
Initialize the selected SDK with the correct key type and evaluate a feature flag. FlagPilot resolves flags from the local cache first, with automatic fallback to the REST API.
import { FlagPilot } from '@flagpilot/node';
const flagpilot = new FlagPilot({
apiKey: process.env.FLAGPILOT_SECRET_KEY,
environment: 'production'
});
const isEnabled = await flagpilot.evaluate('my-flag', {
userId: user.id
});Key features
Local cache evaluation
Flags are resolved from an in-process cache with background polling refresh. No network latency at evaluation time.
AI-native MCP server
A built-in Model Context Protocol server lets Cursor and Claude toggle and inspect flags with natural language.
Real-time analytics
Track evaluation counts, ON/OFF ratios, and audit history for every flag from the dashboard.
Targeted rollouts
Gate features to specific user IDs for internal dogfooding, then graduate to percentage-based rollouts.