Core Concepts

Governance Gate

A structural enforcement mechanism that intercepts AI agent actions before they execute, checks them against institutional constraints, and blocks or escalates violations in real time.

The governance gate is the core enforcement mechanism for AI agent governance. It operates as a PreToolUse hook in the Model Context Protocol (MCP), meaning it intercepts every tool call an AI agent makes before the tool executes.

When an AI agent attempts an action — pushing code, sending an email, creating a file, calling an API — the governance gate: 1. Identifies the action and its context 2. Checks it against active institutional constraints 3. If allowed: the action proceeds and a governance trace is recorded 4. If blocked: the action is prevented, an escalation is created, and a human is notified

This is structural enforcement. The agent cannot bypass the gate because it operates at the infrastructure level, not the behavioural level. It's not an instruction the agent follows — it's a system the agent operates within.

The governance gate is what makes moment-of-action enforcement possible for AI agents. Without it, AI governance is purely behavioural (system prompts, guidelines) and fundamentally unreliable.

How Constellation handles this

Constellation's governance gate integrates via MCP hooks. It works with Claude Code, GPT, and any MCP-compatible AI agent, providing structural enforcement regardless of which AI model is used.

Frequently Asked Questions

Can AI agents bypass the governance gate?

No. The governance gate operates at the infrastructure level (MCP PreToolUse hook), not at the behavioural level (system prompt). The agent cannot execute a tool call without the gate evaluating it first. This is what makes it structural enforcement rather than behavioural guidance.

Does the governance gate slow down AI agents?

Constraint evaluation typically adds milliseconds to each action. This is negligible compared to the time saved by eliminating human-in-the-loop approval for routine actions that fall within established constraints.