Fondoo is built so AI agents can run a business alongside you. This policy explains how agent actions work, what agents can't do, and who is responsible for what. It is part of our Terms of Service.
1. Bring your own agent
- You connect your own agent (for example ChatGPT, Claude, Claude Code, Codex or Cursor) through the MCP connection or the REST API, or you add your own AI key.
- Fondoo does not run AI models on its own account at launch. There's no Fondoo-paid AI, and no AI quota to buy.
- Your agent's provider is your choice and works under your agreement with them. Whatever your agent reads through Fondoo (for example a support ticket or a contact list) goes to that provider. Make sure your own privacy notice covers it.
- We don't use your data to train AI models.
2. How agent actions work
Same powers as the dashboard. Anything you can do in the dashboard, an agent can do through the MCP connection or API, within the permissions of the key or connection you gave it.
Preview first. Write actions can be run as a preview (dry-run) that shows what would change before anything happens.
Approval for risky actions. These need a human to approve, through an approval link or the dashboard:
- deleting things (sites, pages, contacts, connections and other destructive actions);
- spending money (domain purchases and other card-on-file charges, which also respect the spending caps you set);
- emailing a list (newsletters and bulk sends);
- creating new Fondoo API keys;
- other sensitive actions we mark as approval-required, such as acting as one of your app's users in production.
Paid X link posts show the charge before posting, and the agent must send an explicit confirmation. There are no silent charges.
Human-only steps. Agents can see the status of these and send you a link, but only a human can complete them: identity checks (KYC), entering card details, approving OAuth connections, and setting spending caps.
Audit log. Every agent and API action is recorded in your workspace: which agent or key, what it did, when, and the result. You can review and export it.
3. Agents never move money
Agents connected to Fondoo cannot issue refunds, send payouts, or move money in any way. Refunds and payouts are done by a human in Stripe. Support tickets about refunds show a "Handle in Stripe" button for you.
4. You're responsible for what your agent does
- Actions your agent takes are your actions. That includes every page, post, email, product listing, price and support reply it publishes or sends.
- Agents make mistakes. They can write things that are false, outdated, defamatory, infringing or against a platform's rules. Review what matters before approving it, and set approval rules that fit your risk.
- Fondoo runs some automatic checks on supplied content (for example duplicates, broken links and sources, prohibited claims, and page structure). If you add your own AI key, an extra AI quality check can run. These checks reduce risk but don't guarantee accuracy or legality.
- Your agent must follow our Acceptable Use Policy. Using an agent to get around caps, approvals, suppression lists or safety checks is a violation.
- Keep your agent connections and API keys secure, and revoke any you no longer use.
5. Labelling AI content
- Support replies. When a support reply is drafted or sent by AI, Fondoo labels it as AI. Don't remove that label.
- Chat widgets and assistants on your site that use AI must say they are AI. Some laws (for example California's bot disclosure law) require this when bots talk to people to sell something.
- Social networks. Many networks (such as TikTok, YouTube, Instagram and Facebook) require labels on realistic AI-generated or AI-altered media. Follow their rules.
- Reviews and testimonials must come from real customers. Don't publish AI-written reviews or endorsements as if real people wrote them. (The US FTC prohibits fake reviews, including AI-generated ones.)
- Affiliate and sponsored content needs the disclosures the law and networks require, whether a person or an agent wrote it.
6. Escalation to humans
In Fondoo Support, complaints, refund requests, legal issues and threats always go to a human. Agents can draft, but don't use automation to close or answer those without a person reviewing them.
7. Platform rules your agent must follow
- Email: only to people who expect it. Double opt-in for newsletters, one-click unsubscribe, automatic suppression of bounces and complaints. No cold outreach and no purchased lists.
- Phone and SMS: no cold SMS and no AI phone calls.
- Social: posting caps per network; Reddit and Lemmy posts only after a human approves each one.
- Spending: within the caps you set, with approval links.
8. If something goes wrong
- Many actions can be undone from the version history or the audit log view.
- Revoke an agent's access by deleting its connection or API key in the dashboard.
- Report a problem or suspected misuse to legal@getfondoo.com.
We'll update this policy as agent features change, with notice for material changes as described in our Terms.