I have created a vector database with a MCP wrapper as a shared memory for my projects. It works like a charm. Also, I have an agent to agent communication challew with MCP tools for agents. Agents know how to communicate with each other and definately this makes my development a lot easier. Especially communication between repos is valuable. One agent developing IOS app requesting api change from another agent in a backend repo - that works but….
- A reliable way to give an agent-id to an agent
- How to make this agent to check a2a messages with specified intervals
- How to make the agent to act on the request from another agent thru a2a without my intervention
The good thing is that an agent can send a request using a2a without my telling so but there is still so much I would like not to do.
Anybody have any great ideas! Would love to hear! Also, if you have questions or comments would be nice.
PS. Setting these up was not my idea got it from somebody in as a comment to my post in a ai themed facebook group. Took me two mornings to set these great tools up.