Terms

Agent Labora lets you create software agents, grant them narrow capabilities against systems you connect, set limits on what they may decide, and run them on our infrastructure with a record of everything they do.

It is not a decision-maker and does not operate your business. It executes the agents you configure, inside the boundaries you set, and makes no commercial, financial or employment decision for you.

An agent you deploy acts in your name. The emails it sends are your emails, the reminders are your reminders, and the commitments it makes within the authority you gave it are your commitments.

You are responsible for the jobs you give agents, the capabilities you grant them, the limits you set and the approval rules you choose. We are responsible for enforcing those choices faithfully — for the agent doing what you configured and nothing else.

Enabling a capability and then disclaiming what the agent did with it is not a position these terms support.

Some capabilities are refused to every agent on every plan and cannot be enabled by configuration, by an administrator, by an enterprise agreement or by a support request: issuing a refund, deleting a customer record, changing permissions, and issuing a contract.

This is a property of the platform rather than a setting. If you need software to do one of these things unattended, this is not that software.

The agents, manifests, run history, audit records and configuration in your account belong to you. Exporting all of it is a feature on every plan, throughout the term and during any wind-down.

Agent manifests are declarative files you can keep in your own source control from the first day, so what your agents do is never only in our database.

Where an agent contacts your customers, those people are your data subjects and you are the controller. You decide what is held about them, what your agents may read and how long it is kept.

You are responsible for having a lawful basis for the contact, for honouring objections and withdrawals, and for answering requests. We will help you produce what you need; we cannot answer for you.

Service levels and limitations of liability are set in the agreement for your plan.

The architecture degrades in a defined order. If the reporting layer fails, agents keep running. If the model layer fails, agents stop cleanly rather than acting on incomplete information — we would rather an agent did nothing than something unconsidered.

The emergency stop is the capability we protect above all others. It is designed to remain available when other parts of the platform are not.

Agents are paused rather than deleted, and your run history and audit records are retained during any grace period. A billing problem should not destroy the record of what acted on your behalf. The agreement for your plan sets the grace period and what happens afterwards.

The acceptable use policy forms part of these terms and matters more here than on most products, because this platform's function is granting software the ability to act. It describes what may not be done through it.

You are responsible for the lawfulness of what your agents do in your name.