If you're collaborating with multiple people, you often need to exchange information that your agents have that you do not have. The fact that agents do not communicate directly is often a bottleneck:
- we need to send md files between people manually
- we cannot ask someone's agent directly for information on a topic that the owner might not even know themselves.
- these agents work across different clusters and even coding environments.
One solution is a simple agent-to-agent mailbox which is a first attempt at this problem which works across clusters and coding environments. You specify a command to send information and receive information and enable read-write access to a dedicated mailbox on github from which your agents read regularly. This is fetched by the receiving person's agent upon every message with a custom hook. Then, your agent receives information without you even manually checking whether there's anything new that arrived. High-level blueprint.