Repolode
Submit
sipyourdrink-ltd

sipyourdrink-ltd/bernstein

Open-source governance layer for AI coding and other agents: declarative policies, a deterministic scheduler, and offline-verifiable run receipts.

View on GitHubWebsite
Stars
1,194
Forks
185
Open issues
271
Last commit
3 hours ago

What is it?

Bernstein runs multiple AI agents in parallel under rules you write as code. Policies cover who may do what, which actions need approval, and what must be recorded. The coordination loop is plain Python with no model involved, so the same plan produces the same task graph and a run can be checked step by step after the fact.

Coding tasks each get their own git worktree and pass merge gates before landing on the main branch. Tasks can also declare an artifact contract such as a report, dataset or action log, and complete on a signed lineage receipt instead of a commit. Run receipts bind the journal head and optional HMAC-chained audit records under an Ed25519 signature, and can be verified offline from the file alone.

It works with more than 50 CLI agent adapters, including Claude Code, Codex CLI, Gemini CLI, Aider and Ollama, and offers a generic wrapper for any tool with a --prompt flag. It includes a terminal dashboard, a browser dashboard, workflow DAGs in YAML, and an air-gap install profile. It is Apache-2.0 licensed and currently in beta.

Key features

  • 01Declarative policy rules for agent permissions, approvals and recording
  • 02Deterministic Python scheduler with no LLM in the coordination loop
  • 03Per-task git worktrees with merge gates for coding work
  • 04Offline-verifiable run receipts and optional HMAC-chained audit log
  • 05More than 50 CLI agent adapters plus a generic --prompt wrapper
  • 06Terminal and browser dashboards, plus an air-gap install profile

Use cases

  • Running several coding agents in parallel on one repository without conflicting merges
  • Producing audit evidence packs where each step must be reproducible and verifiable
  • Coordinating non-code agent work such as research reports or dataset generation
  • Operating agents in air-gapped environments with a local wheelhouse install

Quick start

  1. 1

    Install Bernstein with uv

    uv tool install bernstein
  2. 2

    Initialize Bernstein in your project

    bernstein init
  3. 3

    Check that a CLI agent is installed and authenticated

    bernstein doctor
  4. 4

    Run a first task with a natural-language prompt

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

Things to know

  • The project is in beta and solo-maintained, and minor versions may change interfaces.
  • It needs at least one CLI coding agent installed and authenticated, such as Claude Code or Codex.

In collections

Multi-agent orchestration frameworksOpen-source AI agent frameworks & apps

Topics

#ai-agents#agent-orchestrator#governance#audit-log#git-worktree#deterministic-replay#llm-orchestration#compliance#agent-fleet#agent-governance

FAQ

What is bernstein?
Open-source governance layer for AI coding and other agents: declarative policies, a deterministic scheduler, and offline-verifiable run receipts. Bernstein runs multiple AI agents in parallel under rules you write as code. Policies cover who may do what, which actions need approval, and what must be recorded. The coordination loop is plain Python with no model involved, so the same plan produces the same task graph and a run can be checked step by step after the fact.
Is bernstein free to use? What license does it have?
bernstein is open source under the Apache-2.0 license. Check the license terms before using it in commercial products.
Is bernstein still maintained?
Actively maintained. The last commit to bernstein was on Nov 7, 2026.
What can I use bernstein for?
Running several coding agents in parallel on one repository without conflicting merges Producing audit evidence packs where each step must be reproducible and verifiable Coordinating non-code agent work such as research reports or dataset generation Operating agents in air-gapped environments with a local wheelhouse install
How do I install bernstein?
Install Bernstein with uv: uv tool install bernstein Initialize Bernstein in your project: bernstein init Check that a CLI agent is installed and authenticated: bernstein doctor Run a first task with a natural-language prompt: bernstein -g "fix the failing test in tests/test_foo.py"
Where can I get bernstein?
The source code and installation instructions are on GitHub: https://github.com/sipyourdrink-ltd/bernstein

Summary generated by AI from the project's README. GitHub stats refresh daily.

Related projects