Looper

Looper does not replace the intelligent agents you are currently using; its role is to connect Codex, OpenCode, Grok, Hermes, and local intelligent agents together.

How to use it

On the canvas, create a diagram in which each node is responsible for only one specific task. Select the appropriate agents, models, tools, workspaces, and backup plans in case of failure, and then connect these elements to define the tasks that need to be carried out sequentially. This entire diagram can be run in a loop, or it can be initiated manually at set times.

During runtime, it is possible to view the messages, tool calls, retry attempts, and execution times of each agent along the timeline. The files are saved in an authorized workspace, allowing them to be used by subsequent nodes and in the next round of operations. Self-hosting is also supported.

Who is it suitable for?

A team that has already been using several programming agents simultaneously and is tired of dealing with overly long prompt strings. For those who simply use one agent to write scripts, such complex tools are completely unnecessary.