Hello, does anybody use any hooks/mcp servers or any similar system for memory within a project?
I would like the agent keep track of what it has discovered during its work, anything it considers useful etc. I had an idea at the end of the chat the agent would analyze what it discovered in terms of the project analysis and possibly save it to the memory.
I just created a shared vector memory for all of my projects. Works really well (auto agents do not allways remember to call memory). The memory is a vector database with postgres sql and a MCP wrapper.
I also set up agent to agent communication which really helps me when project extends in more than one repo.