Levered Docs

Levered Documentation

Infrastructure for self-optimizing products

Levered is a data-warehouse native optimization platform designed for coding agents. It equips your agent with the infrastructure to run Bayesian bandits for real-time product optimization and personalization.

Why Levered?

Most experimentation tools are built for humans to run manual tests. Levered is built for a different model: AI-native products that optimize themselves, continuously, at runtime.

Levered uses data-efficient learning algorithms to dynamically shift traffic toward better-performing variants in real time, while continuously exploring new alternatives. Every user gets the variant most likely to work for them based on their context.

Every capability in Levered is accessible through the CLI and API with structured JSON output, making it easy for AI agents like Claude Code to generate variations, set up optimizations, connect data sources, and monitor results autonomously.

How it works

  1. Connect your warehouse -- Point Levered at your BigQuery or Snowflake instance where exposure and conversion events live.
  2. Define metrics -- Write SQL queries that tell Levered what counts as a reward (conversion, revenue, engagement).
  3. Create an optimization -- Define the design features (headline text, CTA color, layout) and their levels.
  4. Integrate the SDK -- Use the JavaScript SDK or REST API to serve the best variant to each user.
  5. Levered optimizes -- The bandit model trains on your warehouse data and continuously improves variant selection.

Claude Code plugin

The fastest way to get started. Install the plugin and tell Claude what you want to optimize -- it handles warehouse setup, optimization creation, and SDK integration for you.

/plugin marketplace add levered-hq/claude-plugin
/plugin install levered@levered

Then just describe what you want:

> Optimize the paywall to increase free trial starts