We have been playing with repeatable and re-usable flows that someone can run on either Codex, Claude, Cursor or other AI agent. Today we are releasing an open source protocol and a set of tools allowing anyone to declare AI-driven flows (a.k.a. Recipe) and execute them.
In a nutshell, it allows to describe some AI-flow that can be executed by any AI coding agent (Claude, Codex, Cursor). The flow may contain of multiple context files, commands, permissions configuration, etc. The tool generates a setup native to the AI agent, so it works as a provisioning system, not a wrapper, MCP or some sort of other external dependency once provisioned.
Hi all,
We have been playing with repeatable and re-usable flows that someone can run on either Codex, Claude, Cursor or other AI agent. Today we are releasing an open source protocol and a set of tools allowing anyone to declare AI-driven flows (a.k.a. Recipe) and execute them.
In a nutshell, it allows to describe some AI-flow that can be executed by any AI coding agent (Claude, Codex, Cursor). The flow may contain of multiple context files, commands, permissions configuration, etc. The tool generates a setup native to the AI agent, so it works as a provisioning system, not a wrapper, MCP or some sort of other external dependency once provisioned.
Woul love to hear your feedback!