Bernstein

Bernstein is a powerful Python-based orchestrator designed to facilitate the deterministic scheduling of multiple CLI AI coding agents (such as Claude Code, Codex, and Gemini CLI), allowing them to work in parallel across isolated git worktrees. This project enhances productivity by streamlining the coding process, enabling users to specify goals and manage tasks effortlessly. Each operation undergoes comprehensive tracking via an HMAC-signed audit chain to ensure accountability and reproducibility of results.

Key Features

• Multi-Agent Execution: Bernstein can coordinate up to 40+ different coding agents simultaneously, eliminating race conditions through isolated git worktrees. This parallel execution increases efficiency by allowing tasks to be split and addressed concurrently by different agents.
• Deterministic Scheduling: Each run is deterministic, ensuring that results can be replicated and verified. This feature is essential for teams working in regulated environments or those requiring a tamper-evident audit log.
• Ease of Use: Set up Bernstein with minimal effort using package management tools like pipx. The simple command structure allows users to execute complex tasks with straightforward commands, such as:

bernstein run -g "fix the failing test in tests/test_foo.py"

• Compatibility with CLI Agents: Bernstein seamlessly integrates with a variety of CLI coding agents, making it adaptable to existing workflows and allowing for both local and cloud-based solutions. Users can choose from a wide range of supported agents, enhancing flexibility in deployment scenarios.
• Audit Trail: Every decision in the orchestration process is documented in an HMAC-signed audit log, maintaining transparency in the operations performed and ensuring that there is a clear historical record of task executions.
• CI Integration: Bernstein can function as a continuous integration tool, autonomously monitoring pull requests and spawning fixer agents to rectify build failures automatically.
• User-Centric Applications: This tool is specifically tailored for engineering teams that manage multiple coding agents, developers needing effective parallel processing, and organizations that require stringent compliance with regulatory standards.

Use Cases

1. Engineering Teams: Teams running multiple CLI coding agents can leverage Bernstein for efficient concurrency, reducing merge conflicts and maximizing productivity.
2. Regulated Environments: For enterprise applications, Bernstein’s audit logging features provide the necessary compliance documentation.
3. Forward-Deployed Engineering: Utilize Bernstein for client-based work where security is paramount, ensuring that sensitive credentials are managed in a controlled environment.
4. Cost Management: Organizations spending over $1,000 monthly on coding assistance tools can see cost savings through the efficiencies gained with deterministic execution plans.

Conclusion

Bernstein transforms the interaction with AI coding agents by offering deterministic, parallel, and manageable processes suited for various development contexts. From maintaining code quality to efficiently allocating tasks among agents, Bernstein empowers users to achieve their coding goals systematically and reliably.