MCP Catalogs
Homebernstein screenshot

bernstein

by sipyourdrink-ltd·375·Score 52

Audit-grade multi-agent orchestration for CLI coding agents with HMAC-chained audit logs and signed agent cards.

developer-toolsai-llmsecurity
38
Forks
11
Open issues
this month
Last commit
2d ago
Indexed

Overview

Bernstein is a deterministic Python scheduler that orchestrates multiple CLI coding agents (Claude Code, Codex, Gemini CLI, etc.) in parallel git worktrees. It provides HMAC-SHA256 audit chains, signed agent cards, per-artifact lineage, and air-gap deploy capabilities. The scheduler uses plain Python rather than LLMs for task coordination, ensuring determinism and reproducibility while maintaining security for regulated environments.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

you:Engineering teams running 3+ CLI coding agents in parallel
you:Regulated or on-prem environments requiring audit trails
you:Platform teams needing accountability for AI agent decisions
you:How is Bernstein different from other agent orchestrators?
you:Does Bernstein support multiple AI providers?

When to choose this

Choose Bernstein when you need audit trails for AI-driven coding workflows, especially in regulated environments requiring compliance with standards like DORA, NIS2, or EU AI Act Article 12.

When NOT to choose this

Don't choose Bernstein if you need a SaaS solution with vendor support, if your use case doesn't require audit capabilities, or if you're building non-coding LLM workflows.

Tools this server exposes

6 tools extracted from the README
  • idea-to-pr

    Runs a workflow from research to PR implementation

  • refactor-with-tests

    Workflow for finding, proposing, and implementing refactoring

  • security-review

    Workflow for scanning, triaging, and patching security issues

  • doc-update

    Workflow for auditing and updating documentation

  • dependency-bump

    Workflow for bumping and testing dependencies

  • hot-fix

    Workflow for reproducing and fixing critical issues

Note: Tool names inferred from workflow examples in the documentation, though these appear to be workflow configurations rather than direct MCP tools

Comparable tools

claude-flowarchoncrewaiautogen

Installation

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

For MCP server mode:

  1. Install Bernstein via pipx as above
  2. Configure in your MCP client:
{
  "mcpServers": {
    "bernstein": {
      "command": "bernstein",
      "args": ["mcp"]
    }
  }
}

Alternative installation methods:

curl -fsSL https://bernstein.run/install.sh | sh  # macOS/Linux
irm https://bernstein.run/install.ps1 | iex     # Windows
pip install bernstein                           # pip
uv tool install bernstein                       # uv
brew tap chernistry/tap && brew install bernstein # Homebrew

FAQ

How is Bernstein different from other agent orchestrators?
Bernstein uses deterministic Python scheduling instead of LLM coordination, provides HMAC-chained audit logs, signed agent cards, per-artifact lineage, and is designed for on-prem deployment with air-gap capability - features lacking in most other orchestrators.
Does Bernstein support multiple AI providers?
Yes, Bernstein supports 44 CLI agent adapters including Claude Code, Codex, Gemini CLI, and 40+ others. It can wrap any CLI coding tool rather than being limited to specific AI providers.

On Hacker News

Recent discussion from the developer community.

Compare bernstein with

GitHub →

Last updated · Auto-generated from public README + GitHub signals.