In Memoria runs locally (npx in-memoria server) and learns your naming conventions, architectural patterns, and project structure by parsing your code with tree-sitter. It supports 12 languages now, including TypeScript, Python, Rust, Go, and C++.
The AI tools connect over MCP and use that memory to give context-aware suggestions without starting fresh every session.
Would love to hear if anyone else is working on something like this or has feedback on using pattern learning versus explicit configs.
No comments yet.