# Bernstein

> Open-source multi-agent orchestration system for AI coding agents. Decomposes goals into parallel tasks, routes to optimal models, verifies via quality gates, merges results. 21 adapters, Cloudflare cloud execution, deterministic Python scheduling. Apache 2.0.

## Docs

- [Documentation](https://bernstein.readthedocs.io/): Full technical documentation
- [Architecture](https://bernstein.readthedocs.io/en/latest/ARCHITECTURE/): System architecture and design
- [Getting Started](https://bernstein.readthedocs.io/en/latest/GETTING_STARTED/): Installation and quickstart
- [Adapter Guide](https://bernstein.readthedocs.io/en/latest/ADAPTER_GUIDE/): Supported agents and how to add your own
- [API Reference](https://bernstein.readthedocs.io/en/latest/openapi-reference/): Task server REST API
- [Cloudflare Guide](https://bernstein.readthedocs.io/en/latest/cloudflare-overview/): Cloud execution on Cloudflare Workers
- [Configuration](https://bernstein.readthedocs.io/en/latest/CONFIG/): bernstein.yaml reference

## Resources

- [GitHub](https://github.com/chernistry/bernstein): Source code and issues
- [PyPI](https://pypi.org/project/bernstein/): Python package
- [npm](https://www.npmjs.com/package/bernstein-orchestrator): Node.js wrapper
- [Full Technical Reference](https://bernstein.run/llms-full.txt): Comprehensive 500+ line reference for LLMs
- [OpenAPI Spec](https://bernstein.run/openapi.yaml): REST API specification
- [RSS Feed](https://bernstein.run/rss.xml): Blog updates

## Blog

- [Getting Started: Your First Multi-Agent Run in 5 Minutes](https://bernstein.run/blog/getting-started): Install Bernstein, configure Claude Code or any CLI agent, and run your first parallel multi-agent coding session in under 5 minutes. Step-by-step tutorial with examples.
- [Community Spotlight: April 2026](https://bernstein.run/blog/community-spotlight-april-2026): First community spotlight — highlighting contributors who shaped Bernstein's architecture decomposition, adapter ecosystem, and cost-aware routing in April 2026.
- [Running AI Agents on Cloudflare: Workers, Workflows, and Durable Objects](https://bernstein.run/blog/cloudflare-cloud-execution): Bernstein v1.7.3 adds Cloudflare cloud execution — run coding agents on Workers, orchestrate with Durable Workflows, store artifacts in R2.
- [From 4,000 Lines to 200: Decomposing Bernstein's Core](https://bernstein.run/blog/module-decomposition): How 11 parallel agents decomposed a 4,198-line orchestrator into 15 focused sub-packages — using Bernstein to refactor itself.
- [How Bernstein Routes Tasks to the Right Model (and Saves 50-60%)](https://bernstein.run/blog/cost-aware-routing): Bernstein's epsilon-greedy bandit router learns which model fits each task type, cutting agent costs by 50-60% without sacrificing quality.
- [Introducing Bernstein: Multi-Agent Orchestration for CLI Coding Agents](https://bernstein.run/blog/introducing-bernstein): Bernstein v1.0 ships — an open-source orchestrator that runs Claude Code, Codex, and Gemini CLI in parallel with deterministic scheduling and git worktree isolation.

## Optional

- [Changelog](https://bernstein.readthedocs.io/en/latest/CHANGELOG/): Version history
- [Contributing](https://bernstein.readthedocs.io/en/latest/CONTRIBUTING/): How to contribute
- [License](https://github.com/chernistry/bernstein/blob/main/LICENSE): Apache 2.0