Acceptable use policy
AgentGate connects AI assistants to an organization's systems and controls what they can do. This policy explains how it may and may not be used, to protect people, data and the platform. It applies to those who subscribe to AgentGate, to their users and to any agent or assistant acting on their behalf.
Last updated: November 1, 2026
1.Permitted use
- Connecting your own systems, or systems you are expressly authorized to use.
- Defining tools with the minimum permissions each task needs.
- Using AI models and MCP servers for your organization's legitimate tasks.
- Trying and evaluating the platform with data you have the right to use.
2.Prohibited use
You may not use AgentGate, or allow an agent or assistant to use it, to:
- Carry out illegal or fraudulent activities, or activities that harm other people.
- Access systems, data or accounts without authorization, or try to bypass the platform's permissions, confirmations, caps, budgets or records.
- Share keys, credentials or sessions with anyone other than their holder, or use another person's.
- Connect third parties' personal or confidential data without a legal basis or without permission to process it.
- Create or spread malware, or carry out intrusion, espionage, harassment, impersonation or deception.
- Send to an AI model information that the law, a contract or an internal policy forbids sharing with an outside provider.
- Make decisions with legal or serious effects on people, such as credit, employment, health or safety, without a person reviewing and approving them.
- Turn off or skip the confirmation a tool requires before making a real change.
- Overload or interrupt the service, or test its security without a prior written agreement.
- Copy, reverse engineer or resell the platform without an agreement.
3.Responsibilities of the customer
- Deciding which systems, data and tools are exposed, and to whom.
- Keeping access up to date, removing it from people who leave the organization and revoking keys that are compromised.
- Having the legal basis and notices needed for the personal data you connect.
- Reviewing AI-generated tools and proposals before publishing them.
- Complying with the rules that apply to you, such as data protection and industry regulations.
4.AI answers
A model's answers can be wrong, incomplete or unexpected. AgentGate applies controls, such as read-only by default, confirmation of changes, caps and records, but the final decision and the responsibility for using an answer or running an action rest with people. The content of the documents and data that are processed is treated as untrusted, and nothing generated is published unless a person chooses it.
5.Records and oversight
AgentGate records who called, from which client, which tool or model and with what result. Those records are used for auditing and for investigating misuse. Depending on the configuration you choose, content is processed in your network and only metadata reaches the platform.
6.If something goes wrong
If you detect misuse, an exposed key, access that should not exist or a security flaw, tell us as soon as possible through the contact page, with what happened and when. In the meantime, revoke the affected key or access. Please do not publish the details until we have been able to fix it.
7.What can happen after misuse
We act according to severity. If there is an immediate risk, we may suspend access without prior notice.
- A warning and a request to correct the use.
- Suspension of keys, tools or access.
- Suspension or closure of the account, as agreed in the contract.
- Cooperation with the authorities where the law requires it.
8.Changes to this policy
We may update this policy. The date of the last update appears at the top of the page and, if a change is important, we will let account holders know.