Run Bernstein
Bernstein is the open-source governance layer for AI agents; CLI coding agents work out of the box (Claude Code, Codex, Gemini CLI, and 40+ more). There is no hosted instance: the orchestrator runs on the machine where it is installed. This skill covers getting from nothing to a supervised run.
Install
pipx install bernstein
(or uv tool install bernstein). Air-gapped hosts can use the offline
wheelhouse profile — see the install docs.
First run
cd your-repo
bernstein init
bernstein run "add input validation to the checkout form"
init writes the project seed; run plans the goal into tasks and
spawns agents in per-task git worktrees. No model sits in the
coordination loop, so a rerun of the same plan replays byte-identically
and diverging runs are detectable by hash.
Watch and control
bernstein status # task states, one line each
bernstein live # dashboard TUI
bernstein stop # halt the run
MCP access
The package ships an MCP server (stdio transport) so an agent can drive runs through tools instead of a shell:
bernstein mcp serve
Tool surface and card: https://bernstein.run/.well-known/mcp/server-card.json
Where the docs are
- https://bernstein.readthedocs.io/ — full documentation
- https://github.com/sipyourdrink-ltd/bernstein — source, issues
- https://bernstein.run/llms.txt — this site's corpus index