1. Toolports is a connection
Your AI agent, its model and its prompts are chosen and run by you or your model provider. Toolports relays the agent’s requests to the services you connect, within the permissions you set. It does not review whether a request is a good idea.
2. Agents can be wrong
- Agents can misread documents, invent facts, pick the wrong vendor, customer, account or amount, or repeat an action.
- Text inside your data (a vendor memo, an emailed invoice, a customer note) can contain instructions that trick an agent (“prompt injection”).
- Connected Services can return incomplete, stale or wrong data, and agents may act on it.
3. Our controls are aids, not guarantees
Writes off by default, per-tool permissions, high-risk confirm steps, rate limits and audit logs reduce risk. They do not prevent every mistake, and an agent that is allowed to confirm can confirm the wrong thing. Keep writes off until you need them, allow only the tools you need, and keep confirm steps on for payments, deletes and anything you cannot easily undo.
4. You are responsible for agent actions
Every action taken with your keys, including bills, payments, invoices, voids, deletes and emails, is your action. Review agent work before relying on it, and reconcile your books. Toolports is not liable for what your agents do or for output they produce; see the Terms (Disclaimers and Limitation of liability).
5. Not professional advice
Nothing from Toolports or your agents is accounting, tax, legal or financial advice. Ask a qualified professional before relying on it.
6. Your data and AI training
Toolports does not use your data to train AI models. What your agent’s model provider does with data you send it is governed by your agreement with that provider. See the Privacy Policy.