How Claywire's AI agents are allowed to behave.

• Approved-knowledge answers only — agents respond from the workspace's approved knowledge base, project data, and configured disclaimers. They are constrained by a Compliance Guard that prevents answering outside approved material.
• No fabricated claims — agents must not invent prices, possession or delivery dates, approvals, guarantees, or any legal, medical, or financial claims. When information is not in the approved knowledge base, the agent says so and offers a human handoff.
• Brand-voice replies — agents reply in the configured brand tone but remain factual and within approved content.

Agents must hand off to a human for anything sensitive, ambiguous, or outside their approved scope — including negotiation, complaints, legal/medical/financial questions, and explicit requests to speak to a person. Workspaces are responsible for staffing handoff queues so that escalations are answered.

You configure your agents’ knowledge, tone, templates, and workflows, and you are responsible for the content your agents send. You must ensure that knowledge bases are accurate and lawful and that templates comply with the WhatsApp Business Messaging Policy. Claywire provides the tooling and guardrails; the messages remain your communications to your contacts.

• Consent-first — agents may only message contacts who have given valid opt-in consent, consistent with the DPDP Act 2023 and TRAI rules.
• Opt-out enforcement — agents must recognise and honour STOP / opt-out requests immediately and stop further messaging to that contact across automations.
• Frequency caps — message frequency must respect configured caps and reasonable limits so contacts are not over-messaged; automations should not re-trigger against opted-out or unresponsive contacts beyond configured nurture limits.

Agents and automations must not be used to:

• Cold-outreach to non-consented numbers, or to scraped, purchased, or unlawfully obtained lists.
• Send spam, unsolicited bulk messaging, or content designed to evade WhatsApp quality and policy controls.
• Engage in fraud, deception, phishing, impersonation, or harassment.
• Generate hateful, obscene, defamatory, or otherwise unlawful content.
• Provide regulated professional advice (legal, medical, financial) as if from a licensed professional.

Agent actions — messages sent, automations triggered, opt-out events, and human handoffs — are recorded in an audit trail so workspace owners can review behaviour and demonstrate compliance. You should review agent activity regularly and adjust knowledge and guardrails as needed.

Use of agents over WhatsApp is subject to the WhatsApp Business Messaging Policy and Meta’s terms, including approved templates for business-initiated messages, respect for messaging limits and quality ratings, and prohibited-content rules. Repeated policy violations may result in template rejection, number restriction, or account suspension by Meta or by Claywire.

We may suspend or restrict agents or workspaces that violate this policy, the Terms, or applicable law, or that threaten platform quality or Meta compliance. This policy supplements the Terms & Conditions and Privacy Policy.

Questions about agent behaviour or this policy: