Bernstein, developed by Sipyourdrink Ltd, is a free MCP server tool built with Python. It functions as a deterministic multi-agent orchestrator, connecting to over 40 Command Line Interface (CLI) coding agents such as Claude Code, Codex, Gemini CLI, Aider, and Cursor. Distributed as a Python package, Bernstein is installed using "pipx install bernstein" or "pip install bernstein". Its primary role is to coordinate tasks across these agents, enabling complex workflows for individual developers and teams. It supports standard MCP transport protocols including stdio, HTTP, and Server-Sent Events.

What does Bernstein connect to?

Bernstein connects to various CLI coding agents, acting as a deterministic multi-agent orchestrator. It supports Claude Code, Codex, Gemini CLI, Aider, Cursor, and more than 40 other agents. Its category is an MCP server and a multi-agent orchestrator. Compatible AI clients include Claude Code, Claude Desktop, Cursor, Gemini CLI, and any MCP client, utilizing standard MCP transport protocols like stdio, HTTP, and Server-Sent Events.

Core features and tools

Bernstein orchestrates tasks across other CLI coding agents. It does not expose distinct functional tools like read_file or get_weather for direct AI client invocation via the MCP interface. Instead, it offers internal CLI commands for management: "bernstein verify receipt", "bernstein bench run", "bernstein bench reliability-verify", "bernstein init", "bernstein doctor", "bernstein -g", and "bernstein integrations list". These commands manage Bernstein itself or its integrations, not callable tools for AI clients.

Who is it for?

Bernstein serves individual developers, teams building agentic systems, and those working in compliance-sensitive or production-critical paths. It provides capabilities for accountability and deterministic operations. Its initial design considers a single developer on a laptop. It supports an opt-in HMAC-chained audit log for offline verification. Skip this if your primary need is a server exposing a collection of distinct, directly callable tools to an AI client.

How to get started

No credentials are required for its MCP server mode by default. For auditability, an opt-in HMAC-chained audit log can be enabled by setting BERNSTEIN_AUDIT=1.

Use cases

• Orchestrating multiple CLI coding agents in parallel to decompose, spawn, verify, and merge code changes in isolated git worktrees.
• Enabling deterministic replay of past runs and offline verification of audit logs for accountability.
• Running agents on cloud runtimes like Cloudflare Workers with R2 workspace synchronization.

Bernstein provides a powerful orchestration layer for numerous CLI coding agents, offering deterministic execution and auditability features. It is well-suited for developers needing to coordinate complex agentic workflows and maintain accountability. Its design prioritizes operational control and verification, making it a valuable tool for specific development needs.