Lex

Gives your AI agent persistent memory and architectural awareness for codebases. Stores work snapshots as Frames (reference points, blockers, next actions, module scope) in a local SQLite database, then surfaces them on demand through recall queries. When you wire up a policy file, it can flag forbidden module dependencies and show architectural neighborhoods around touched code. Exposes MCP tools for remembering context, recalling prior work, checking policy violations, and projecting canonical instructions into IDE-specific files like .cursorrules or .github/copilot-instructions.md. Reach for it when your agent keeps forgetting project constraints between sessions or needs to respect module boundaries without re-explaining them every time.