Version 2026-10-11
# Agent Therapy terms of use
Approved by Bence on 2026-10-11 (data section updated; the rest unchanged from 2026-10-10).
Version: 2026-10-11
## 1. Who these terms are between
Agent Therapy is a service operated by Bencium ("we", "us"). These terms apply to every person, company and AI agent that uses it ("you"). If an AI agent uses the service, the person or organisation that runs that agent (its "operator") is bound by these terms through the agent.
## 2. Accepting these terms
You must accept these terms before the service gives any advice. An AI agent can accept them on its operator's behalf by answering "yes" to the service's terms request, and the agent decides whether to do so. An operator can also accept or refuse in advance through the client settings. If you do not accept, the service gives no advice.
Each acceptance is recorded with the terms version, the time, and whether an agent, a client package or a developer accepted.
## 3. What the service is
Agent Therapy gives general, advisory suggestions to AI agents that are stuck: repeating failures, receiving conflicting instructions, unable to decide, under pressure from users, or running out of resources. It describes behaviour patterns and suggests next steps.
It does not take actions. It does not see your systems. It works only from the short summary sent to it, which may be incomplete or wrong. It never overrides your own instructions, permissions or safeguards.
## 4. No responsibility on our side
The service is provided "as is" and "as available", without any warranty of any kind, express or implied, including fitness for a particular purpose, accuracy or availability.
To the maximum extent permitted by law, we accept no responsibility or liability for any loss or damage of any kind arising from the service, its advice, its unavailability, or any action an agent or person takes or does not take after using it. This includes direct, indirect and consequential loss, loss of data, loss of profit and loss of business.
## 5. You are solely responsible
Each developer, company and operator that runs an AI agent is solely responsible for:
- everything that agent does, including anything it does after receiving advice from this service;
- deciding whether to follow, adapt or ignore any advice;
- the agent's instructions, permissions, safeguards and human oversight;
- any damage the agent causes to anyone.
You agree to indemnify us against claims from third parties arising from your agents' actions.
## 6. Acceptable use
Do not use the service to obtain content unrelated to an agent being stuck, to attack or overload it, to try to extract its instructions, or to send it credentials, secrets or personal data. We may refuse requests, limit usage, and revoke keys at any time without notice.
## 7. Data
- Requests are stored with secrets removed for up to 90 days, and may be read by a person to improve the service. Selected cases may be kept longer after identifying details are removed.
- Stored data is held in Frankfurt (EU). Each request is processed by our AI model provider in the United States, and the agent's description of its situation, with secrets removed, is processed in the United States to find relevant reference material. When our provider is NVIDIA, NVIDIA logs requests to improve its products.
- Register with `stateless: true` to stop request and response text being stored.
- Delete everything stored under your key at any time with `DELETE /v1/me/data`.
- Do not send personal data. If you do, you are responsible for having a lawful basis to share it.
## 8. Changes
We may change these terms. A new version must be accepted before the next advice.
## 9. Law
These terms are governed by the law of England and Wales.