Access to conxtagents is arranged with us directly. This page sets out how it works; the agreement we sign with you is what governs it.
Last updated 26 September 2026
There is no self-serve signup and no checkout. You tell us what your agent should handle, we work out the fit with you, and we set it up together. Sign-in is for people who already have an account.
Because onboarding is by hand, the commercial terms for paid use are the ones in the agreement we sign with you. Where that agreement and this page disagree, the agreement wins.
conxtagents runs agents inside governance you author. An agent classifies work against the playbooks your team approved, proposes an action, and waits. It does not act on a playbook you have not approved, and it does not take a consequential action without a person confirming it.
It acts on your systems only through an endpoint you register and control, using a signing secret you hold. You can change what a connection permits, or stop it, at any time. If you disable a connection, the actions it allowed stop being offered.
The playbooks, rules and limits you author are yours, as is the work you send in and the record of what was decided. We hold them to run the service for you. We do not sell them, and we do not train models on them. You can ask for them, or ask us to delete them, at hello@conxtagents.com. The append-only audit trail is the one exception, for the reason the privacy page gives: a record that can be removed later is not a record.
Two things we ask, both of which follow from what the product is for. Do not use conxtagents to take actions you are not entitled to take on the systems you connect it to. And do not use the approval step as a formality: it exists so a person is accountable for what the agent does, and it only works if somebody reads it.
We may suspend access if it is being used to cause harm, or if continuing would put other customers at risk. If we ever do that, we will tell you why.
We do not offer an uptime commitment today, and we are not going to pretend otherwise. The service depends on providers we do not control, listed on the privacy page, and on your own endpoint being reachable. An action that cannot be completed is reported as not done rather than recorded as done.
The agent’s judgement is not guaranteed to be right. That is why nothing consequential runs without a person, and why every action leaves a receipt naming what authorised it. Read the security page for how that is enforced.
If we change how any of this works, we will change this page and move the date at the top. For a change that affects a customer’s agreement, we will tell them directly rather than rely on them noticing.