Your AI agents act. Who decides what they're allowed to do?
Your platform's permissions say who can touch what. They do not say whether the action is permitted by your regulatory obligations — and that is the question a customer, an auditor or a regulator will ask you.
An advisory ALLOW / APPROVE / DENY / ESCALATE decision before every sensitive action, over REST, MCP or A2A — with signed evidence your counterparty verifies themselves, no account needed.
Work platforms
- monday.commonday.com permissions decide who can touch what inside the platform. They do not decide whether the action itself is permitted by your regulatory obligations — that is the missing layer, and it plugs in with a single HTTP call.restmcp
- Microsoft Copilot StudioCopilot Studio lets anyone build agents without code across the productivity ecosystem. The easier the agent is to create, the more urgent the question: who checked that what the agent shares is allowed to be shared?mcprest
- ServiceNow AI AgentsITSM and HR agents process the most sensitive data a company holds after its customers': its employees'. And automated decisions about employees are exactly what several regimes regulate most strictly.restmcp
CRM and marketing
- Salesforce AgentforceA CRM agent touches the most regulated data a company holds: its customers'. Platform guardrails frame the model; they cannot know whether THIS contact, to THIS customer, in THIS jurisdiction, respects consent and transfer rules.restmcp
- HubSpot BreezeA prospecting agent acts on people who asked for nothing. It is the one case where the legal basis of the contact — not the transfer — is the central question, and it changes with the recipient's country.restmcp
Model providers
- OpenAIAn agent that calls tools decides for itself which ones to call. That is what makes it useful — and why a guardrail it can choose to skip is not a guardrail. The authority decision is called from your code, not from the prompt.mcprest
- Anthropic ClaudeAnthropic proposed the Model Context Protocol, now the agent integration standard. StructureClerk exposes it natively: wiring the authority layer into a Claude agent fits in a configuration file, without writing a line of code.mcprest
Automation
- ZapierEach step of a Zap looks harmless on its own. Chained together, they move personal data from tool to tool — and often from country to country — without anyone assessing the transfer.restmcp
- n8nSelf-hosting n8n keeps the automation engine in your jurisdiction. It says nothing about the jurisdiction of the services it calls — and that is exactly where the transfer happens.restmcp
Three protocols, one decision
Whatever the platform, the decision rendered is the same and signed the same way. Only the access path changes.
- REST
- An HTTP call from any workflow or codebase. The universal path: anywhere your platform can call a URL.
- MCP
- The Model Context Protocol, the agent integration standard. Our server exposes 5 tools over Streamable HTTP — one configuration entry is enough.
- A2A
- The Agent-to-Agent protocol, over JSON-RPC. The authority layer becomes an agent your agents query, discoverable through its public card.
Frequently asked questions about AI agent governance
What is AI agent governance?+
It is deciding, before an agent acts, whether the action it is about to take is permitted — and keeping a verifiable record of that. Platform permissions govern access to data; agent governance covers the action itself, its recipients and its jurisdiction.
Why an external layer rather than the platform's own guardrails?+
Because a platform's guardrails are written, evaluated and attested by that same platform: it is self-attestation, unverifiable from outside. An independent layer renders a signed decision a third party checks themselves — and it covers all your agents, whatever platform hosts them.
How do I produce an audit trail of AI agent actions?+
By rendering a decision before each sensitive action: every call produces Ed25519-signed evidence, timestamped and chained to the previous one, carrying the rule applied and the reason. The evidence verifies publicly, without an account — a trail a third party can rely on, not an internal log.
Does StructureClerk block my agents' actions?+
No, and that is deliberate: StructureClerk decides, your infrastructure enforces. The decision is advisory, which keeps the layer out of your critical path — your workflow chooses whether to stop, request validation or continue.
What does the integration cost?+
The assessment that produces your profile is free. For decisions, the design partner program is 0 $ for six months — the offer built for first integrations. After that, Authority plans bill by decision volume, with prepaid credits beyond the allowance rather than a wall.