Back to projects
USMBA-Cortex poster with archives, a search lens and Rust, TypeScript and MCP annotations
Documented case study
Document search · Agent tooling
USMBA-Cortex — A Local Knowledge Workspace
A Rust core for document retrieval, a TypeScript CLI for Notion and an MCP server for assistants.USMBA-Cortex organises an academic workspace around three components: a Rust core with redb storage and inverted-index search, a TypeScript CLI for Notion synchronisation, and a Node.js MCP server exposing search and academic deadlines to assistants.
- Period:
- August 2026 · public source
- Role
- Rust and TypeScript development, MCP integration
Context
- Structure
- Personal project · academic workspace
- Objective
- Connect course documents, deadlines and assistant tools in a local workspace.
- Role
- Rust and TypeScript development, MCP integration
Features
- Local redb storage, an inverted index and keyword search across documents.
- Revision priorities calculated from exam deadlines and module coefficients.
- TypeScript commands for workspace diagnostics and Notion synchronisation.
- An MCP server calling the Rust binary for search and exam information.
- An invariant test scanning Notion source for deletion and archiving operations.
Inside the source
Local keyword search, connected to assistants through MCP. The generative synthesis module remains a prototype.
- Storage and local search
redb tables, word indexing and document retrieval.
- MCP gateway
Search and exam tools connected to the Rust binary.
- No-delete invariant test
Static checks for prohibited Notion operations.