Your systems, your data, your model account.
The agent is built inside your infrastructure and runs there after handover. This page sets out what we touch, when we touch it, and what we sign before any of it starts.
It lives in your infrastructure.
The agent is deployed in your cloud or on your servers, under your access rules. We don't host it, and we don't keep a copy of it running anywhere else.
The agent calls a language model through your own account with the provider you choose. Your documents go to a provider you already hold terms with, under those terms.
Hosting and model usage are billed to you directly by your providers. You see every line of it.
After handover your team runs it. If you want us to maintain it, that is a separate agreement with its own billing. Nothing stops working if you never sign one.
Access comes in steps, and you can take it back at any point.
Read-only
We watch the process as it runs. We can read what the systems hold. We can't change any of it.
Read-only access for named people on our side, listed in writing.
No new access
Design happens with your process owner. Goals, guardrails and escalation points are agreed before the agent is allowed to act on anything.
The same read-only access. Nothing added.
Write access, on your sign-off
The agent runs in parallel with your team. Every action it takes is logged in your environment.
Write access scoped to the systems this one process touches, granted when you approve it.
Our access ends
Every credential we held is revoked on the handover date, and you confirm it. If you sign a maintenance agreement, the access it needs is written into that agreement.
No credentials and no copies of your data.
What happens to your data.
We may ask for sample documents to test against before the build. They're covered by the NDA, used only for that assessment, and deleted when it ends. We confirm the deletion in writing. If you'd rather samples never leave your systems, we run discovery inside your environment from day one.
The agent reads from your systems and writes back to them. We don't export your data to storage of our own, during the build or after it.
We don't use your data to train anything. Your model provider's terms govern what it does with what it receives. We'll go through those terms with your team before the build starts.
Every input the agent read and every action it took is recorded in your environment. Your auditors read the same log your team does.
What we sign before we start.
Signed before you share a single sample. Your template or ours, whichever your legal team prefers.
Available on request for any engagement that touches personal data.
We fill in your security questionnaire as written. Where a question doesn't apply to a studio our size, the answer says so.
You set the notification window. Tell us your requirement and it goes into the contract as written. If something breaks, you hear what happened, what we're doing about it, and when it will be fixed.
We are not SOC 2 or ISO 27001 certified. Because the agent runs inside your infrastructure, it sits under the controls you already audit. What we can show you is the access we hold, the dates we hold it, and a log of everything the agent did.
Send this page to your security team.
If they have a question it doesn't answer, we'd rather hear it before the build than after.