herd-core 0.0.1

Agent team governance framework — execution-agnostic roles, authority models, decision capture, and craft standards for AI agent teams.

The Herd

Agent team governance framework.

Execution-agnostic roles, authority models, decision capture, and craft standards for AI agent teams.

Your agent framework runs the agents. The Herd runs the team.

Status

Pre-release — package name reserved. First functional release coming soon.

What is The Herd?

The Herd is a governance framework for AI agent teams. It provides:

• Canonical role definitions with authority boundaries
• Craft standards that shape how agents work, not just what they do
• Decision capture (HDR — Herd Decision Records) from natural conversation
• Pluggable execution — bring your own agent engine (Claude Code, Codex, CrewAI, or anything else)
• Team topology — multi-team, multi-machine, multi-engine in one config file
• QA gates with tiered validation (automated + architectural review)

Architecture

The Herd is modular by design:

Package	Purpose
herd-core	Core framework, interfaces, topology, roles, HDR, craft
herd-agent-claude	Execution adapter for Claude Code CLI
herd-store-duckdb	Storage adapter (local DuckDB + MotherDuck)
herd-ticket-linear	Ticket lifecycle adapter (reference implementation)
herd-repo-github	Repository operations (PRs, branches, commits)
herd-notify-slack	Notification output

Quick Start (coming soon)

pip install herd-core herd-agent-claude herd-store-duckdb
herd init

License

MIT