Introduction

What is DOF-MESH?

DOF-MESH (Deterministic Observability Framework) is a pre-execution governance layer for autonomous AI agents. It intercepts every agent decision before it executes, converts it into mathematical constraints, and verifies it with Z3 theorem proving in milliseconds — then registers an immutable keccak256 hash on-chain.

Zero-LLM governance. Every decision the framework makes — blocking policy violations, verifying code safety, proving formal invariants — is computed deterministically using Constitution rules, AST analysis, and Z3.

Most frameworks check what happened. DOF checks what is about to happen.

Key Metrics (v0.8.0)

• Core modules: 173
• Tests: 4,800
• Lines of code: 65,360
• Import errors: 0
• AI providers: 8+
• Security layers: 7
• Formal Z3 proofs: 4/4 PROVEN
• Chains deployed: 9
• On-chain attestations: 30+
• ASR governance: 2.3% (regex) / ~4.5% (multi-layer)

Why it matters

Autonomous agents are beginning to perform actions, not just generate responses. They can write code, call tools, modify files, prepare transactions, interact with APIs, and touch operational workflows.

The next risk is not only a wrong answer. It is an unauthorized action.

DOF-MESH addresses that risk before execution by evaluating proposed actions against deterministic governance rules and producing auditable evidence for each decision.

Decision model

Every agent action is classified before execution:

• ALLOW — the action can continue.
• REVIEW — the action requires human or additional validation.
• BLOCK — the action is stopped before execution.

Current positioning

DOF-MESH is a PMV-stage governance framework for controlled pilots, technical validation, and reproducible sandbox demonstrations.

It should not be described as a certified regulatory solution, a guarantee of legal compliance, or production enterprise infrastructure without further third-party validation.