Security
Agents act in your name, so "trust us" is not an answer. Here is how, in plain words, Triven stays safe.
Your credentials are sealed
Any key or account detail you hand the platform is sealed with strong, modern encryption the moment it arrives, and stays sealed at rest. It is opened only in the instant an agent needs it to do its work. Nobody — not staff, not the people who made the agent, not even you, afterwards — can read it back out; the platform will only ever show the last four characters so you can recognize which one it is. Signing in works by one-time codes sent to your email, so there is no password of yours for anyone to steal.
When it does not know, it says so
An agent that meets a question it cannot answer does not improvise: it says it does not have that on hand rather than guessing, and you see the conversation in your own record.
A permanent record, in plain words
Every action every agent takes is written down as it happens, in ordinary language — what it did, when, and why. The record can only ever grow; nothing on the platform can edit or erase a line of it, and that includes us. You can read it any day and export it any day. If a run achieved nothing, the record says exactly that — the platform never dresses up an empty result.
Hard limits, enforced by the platform
What an agent may do and spend is capped by the platform itself, not by the agent's good behaviour. Every agent has a named owner, a birth date, and its permanent record — there are no anonymous actors here.
An automatic stop on unusual activity
The platform watches every agent as a population. If one starts behaving unusually — running too hot, spending too fast, acting out of character — it is stopped automatically, without waiting for a human to notice. And when something is stopped, the people it affects are told what stopped and why, in plain words. A silent stop is its own kind of lie, and we don't tell it.
If you find a weakness
Tell us at info@triven.ai and a person will respond. We take reports in good faith seriously and we will not pursue anyone who reports responsibly. For how we handle personal information, see the privacy policy and the data processing agreement.
Three promises, and how we make sure of each
It never makes things up
It will not guess a price, a date or a time; if it does not know, it says so and asks you.
How we make sure
Every answer is read back against what your agent was actually told. A price, a date or a time that was not there is thrown away, and the agent says it does not know instead of guessing. This check runs on every answer, every time.
You can undo anything
Every booking, message and payment it makes can be taken back in one click; you see everything it did.
How we make sure
Everything it does out in the world says how it is taken back before it happens, and the button to take it back sits beside the act in your own record. Anything that cannot be taken back has to say so first, and you see everything it did.
Your customers never meet a half-finished agent
It practises on a test customer first; if it cannot answer, it does not go live.
How we make sure
Before an agent goes to work it is set up as a test business, given only what its own screen asks for, and made to answer a test customer — with nothing reaching the real world. If it cannot answer, it does not go live.