Metronisys core governance assets
Metronisys™ is a trademark of John A. Nudd.
Commercial use of the name requires permission.
Open-source repositories may be used freely under their respective licenses.
Autonomous AI agents such as UiPath agents or Clawdbot (now Moltbot) represent a fundamental shift in how software operates.
These agents can execute commands, automate workflows, call tools, and act independently in real-world systems.
With this power comes a critical question:
If AI agents can act autonomously — who governs their behavior in the best interest of the human?
Metronisys™ provides the answer.
The Metronisys Manifesto defines why AI autonomy must be governed.
Metronisys Core defines how those principles are enforced in real systems.
| Manifesto Principle | Core Enforcement Mechanism |
|---|---|
| Human authority over automation | Human-in-the-loop escalation |
| Bounded autonomy | Resource & cost guardrails |
| Transparency over power | Tool & MCP call auditing |
| Accountability across systems | Cross-agent delegation integrity |
This repository translates philosophy into enforceable system boundaries.
Metronisys Core is designed as a governance-first orchestration layer that sits between autonomous agents and the environments they operate in.
It does not replace agents.
It governs how far agents may act.
[ Human Authority ] ↑ [ Metronisys Core ] ↑ [ Autonomous Agents ] ↑ [ Tools • APIs • Systems ]
Each module below maps directly to a Manifesto principle and is documented as an explicit governance capability.
📄 escalation.md
Manifesto alignment:
Autonomy must always remain subordinate to human judgment.
Defines when agent execution must pause and request explicit human approval for high-risk actions.
📄 budgets.md
Manifesto alignment:
Power without limits is not intelligence — it is danger.
Defines enforceable boundaries on token usage, cost, execution time, and resource consumption.
📄 tool_audit.md
Manifesto alignment:
If an action cannot be explained, it cannot be trusted.
Ensures all tool calls and Model Context Protocol (MCP) interactions are observable, auditable, and attributable.
📄 delegation.md
Manifesto alignment:
Responsibility cannot be delegated away.
Defines how authority, context, and accountability must propagate across multi-agent systems — preventing indirect bypass or collusion.
Metronisys Core is intentionally:
- Governance-first, not feature-first
- Vendor-neutral, not platform-bound
- Architecture-driven, not speculative
- Human-aligned, not agent-optimised
This repository documents what must exist for safe autonomy — not how fast it should be shipped.
This repository contains:
- Governance architecture
- Enforcement intent
- Machine-readable policy concepts
It does not claim production-ready implementations.
Metronisys Core defines the minimum conditions under which autonomous AI systems should be allowed to operate.
📜 Next:
- Read the Metronisys Manifesto
- Explore agent-facing rules in
agent-governance