I’ve been creating my own agent memory system, which it composes tasks, updates when completed, partial tasks, writes and marks tests etc
Allows for brainstorming session, so that it can gather all the thoughts/ideas towards how to action a certain project/application/step then create the tasks from it etc.
It documents what it has done, for each file & any dependencies (as in other files that rely on it / it relies on)
Also gave it the ability to query my local Ollama LLM to save on actual calls to cursor for quicker responses (like breaking down & creating the tasks)
Just really wanted to better/extend its memory so it can check if its breaking anything or how things work again when it ‘forgets’
Created a CLI for it so that the agent can easily interact with it (Idea to have context easily over multiple conversations etc)
The tool works, but obviously the agent isn’t coded to be able to properly use the tool, I can prompt it to use it but this all adds time/complexity to this issue.
Is there going to be any custom tool support for this agent? Would be great to just have it automatically call this tool etc
Still can’t get cursorrules etc to actually work & not sure if theres been any documentation added towards that yet so we can better understand it.