MentoGuard

Problem Statement

Autonomous AI agents managing real money introduce a new class of risk: they act faster than you can react, and standard off-chain guardrails — environment variables, middleware checks, rate limits in code — live in the same process as the agent and can be overridden. There is no trustworthy, auditable way to tell an autonomous agent you may never exceed these limits and have that be genuinely enforceable. MentoGuard solves this with on-chain constrained autonomy: delegation rules stored in a smart contract that the agent must read and respect before every execution. The owner controls the rules via on-chain transaction. The agent cannot override them, even under adversarial prompting. Combined with Self Protocol identity verification and a Filecoin audit trail, MentoGuard demonstrates that autonomous agents can be both capable and provably constrained — a pattern applicable to any agent managing real assets.