Bernstein Multi Agent Orchestration

What is Bernstein Multi Agent Orchestration?

Bernstein Multi Agent Orchestration is a control plane for orchestrating AI coding agents. It decomposes high-level development goals into smaller tasks, assigns them in parallel to multiple AI agents running in isolated worktrees, and verifies all outputs with a dedicated janitor agent — enabling reliable, thoroughly tested code without lock-in to a specific agent vendor.

How to use Bernstein Multi Agent Orchestration?

Define a development goal and optional agent configuration in a YAML file, then run Bernstein to orchestrate multiple AI agents (e.g., Claude Code, Codex CLI, Gemini CLI) across providers. The server handles task decomposition, parallel execution, and automated verification automatically.

Key features of Bernstein Multi Agent Orchestration

• Declarative YAML orchestration for complex dev workflows
• Parallel execution of multiple AI agents across providers
• Automated output verification with a dedicated janitor agent
• Vendor-agnostic and portable prompts and workflows
• Continuous self-improvement through automated feedback loops

Use cases of Bernstein Multi Agent Orchestration

• Automate complex dev workflows from a single goal description (new features, refactoring, API docs).
• Embed AI-driven code generation and testing directly into CI/CD pipelines for zero-touch operations.
• Continuously optimize agent coordination and outputs through feedback-driven refinement.

FAQ from Bernstein Multi Agent Orchestration

What is the janitor agent?

The janitor agent is a specialized agent that checks code correctness, runs tests, and prevents regressions after parallel agent execution.

Is Bernstein tied to a specific AI agent vendor?

No, Bernstein is vendor-agnostic and supports multiple providers including Claude Code, Codex CLI, and Gemini CLI. Prompts, workflows, and commands remain portable between them.

How does Bernstein ensure code quality?

Every agent output is automatically verified by the janitor agent, which runs tests (e.g., pytest, eslint) and ensures no regressions are introduced.

Can I define custom workflows in Bernstein?

Yes, you define complex dev tasks and workflows with simple declarative YAML configuration, specifying the goal and the agents to use.