What We Build

The Agent Manifesto gives you a stable foundation for building AI instruction systems that scale with your project — not against it.

A Structured Instruction System

At the core is a root contract (AGENTS.md or a tool-native file) that holds classification rules and capability routing. Skills execute domain work. Pipelines orchestrate multi-step flows. Agents provide context isolation for independent review.

Typical Outcomes by Project Size

• Small — minimal root contract, mandatory protocol-derived skills, no manager by default
• Medium — manager-equivalent routing, at least one pipeline, explicit validation on every non-trivial task
• Large — multiple pipelines, selective agents where context isolation is justified, full reference docs

Compatible With Any AI Tool

Tool-agnostic: Claude Code, Cursor, GitHub Copilot, or any agent that can read files. Single-tool projects use the tool's native entrypoint; multi-tool projects use AGENTS.md as the shared root with thin tool-specific adapters.