Overview
Lending-Mind Protocol makes engineering judgment executable without turning it into invisible prompt context. A Mind Package records the decisions, constraints, and evaluation rules that should govern an AI-assisted change.
Rust owns the protocol semantics, policy evaluation, signing, artifacts, registry operations, daemon, and MCP behavior. Node.js provides delivery and onboarding glue, while the Python orchestrator handles sandboxed qualification and benchmark workflows.
What I built
- Local-first control-plane concepts for engineering judgment
- Signed preference and evaluation packages
- Bounded execution with explicit review evidence
- Rust parsing, policy evaluation, artifacts, registry, daemon, and MCP boundaries
- A foundation for repeatable AI-assisted software decisions
Stack
Rust, Node.js, Python, MCP, signed packages, local persistence, and protocol-oriented tooling.
What I learned
The most useful AI systems are not only more capable; they are more accountable. Encoding preferences, evidence, and approval boundaries makes a system easier to audit and safer to evolve.
