Common questions about how Reynforce works
Does Reynforce block my agent?
No. It keeps a record, and its answer tells your agent so. Every answer carries observation_only: true, which means Reynforce is only watching. An agent that respects this will keep working the same way on the day blocking becomes available. What you get today is your list of facts, a refusal when an agent tries to use a fact nobody stands behind, and a complete record of what went out.
Do you read our email or our CRM?
Only if you connect one, and then only what that connection asks for. A mailbox connection reads sent emails only. It asks for details such as who an email went to and when, and it never asks for the body of the email, so there is nothing to throw away afterwards. A CRM connection asks only for the fields you have chosen.
What do you store from a message?
The sentence a promise was made in, up to 1,000 characters. That is the only text from customer messages that we keep. The limit is written into the format of the record itself, so you can check it for yourself. We do not keep the rest of the message.
Is our data encrypted?
Everything is encrypted while it travels between systems. When it is stored, your data sits on encrypted storage from our hosting provider. That encryption is theirs, with their keys, and we do not claim it as ours. What we encrypt ourselves, with our own keys, is the passwords and keys for any system you connect, and every backup before it is written. /privacy.md explains this in full, including the eight things about it that are not settled yet.
Where does our data live?
In your own separate copy of Reynforce: one application and one storage volume for each customer, in London. We chose this so that a problem with one customer stays with that customer. A login that belongs to one customer is turned away by every other customer's copy.
Can an agent sign itself up and pay?
Yes. But it can never accept a fact or approve its own rules. Facts an agent writes stay proposed until a person accepts them. If an agent checks against one of those, the answer is not yet confirmed by a person, and the agent is not told allowed. Paying changes none of this. A paying agent gets exactly the same tools as one that hasn't paid.
What happens when the agent does not know?
It passes the question to a person and tells the buyer it is checking. There is no setting that lets it answer anyway. Staying silent and giving a confident wrong answer are the two failures the whole product is built to avoid, so neither one can be switched on.
How is this different from guardrails or tools that trace what agents do?
Those tools check what an agent says. Reynforce checks what it commits your company to, and keeps the record. Most teams that come to us already use one of those tools and have just found the gap. /compare.md has the detail, with sources.
What does it cost?
£99 a month for each agent if you do it yourself. The 7-Day Agent Approval, where we do it for you, starts at £2,500, paid once. For an enterprise plan, we agree the price with you on a call. These are early prices and may change. You can cancel any time. It's 14 days free, and no card is needed. /pricing.md.
Who is behind it?
One person, for now. We say so on /founder.md, and we don't hide it.
Is it ready?
It is in early access, and only parts of it are ready. The list of facts, the record, the checks and the agent connection are built and running. No customer has used it for real work yet. There are no case studies, and there is nobody we can put you in touch with. We would rather tell you that than let you think otherwise.