Your data stays yours. Every action is logged.
The platform follows your own processes and permissions. Every agent is isolated, only sees what its role allows and logs every action. Money and promises wait for a person.
Six rules we never bend.
AI that works in your systems needs to be accountable. These rules are built into the platform, not added afterwards.
Yours, not rented
Everything is built around your own processes. You own the code, the prompts, the data and the runbooks.
Every agent isolated
Each agent runs in its own container with its own keys, and can only use the actions it is allowed.
Permissions per role
People and agents only see what their role allows, including in the company brain.
Everything logged
Every read, every action and every approval is in the audit trail.
People approve what matters
Money, promises and public content wait for a person by default.
No training on your data
Models are used through business APIs that do not train on what you send.
Security you can see
Every control, every action and every blocked attempt, live. You decide with us what each agent may read, do alone and only do after approval.
Security and governance
Every agent isolated, every action checked and logged
- Every agent isolatedOwn container and own keys
- Permissions per roleChecked on every request
- Approvals on moneyAbove the limits you set
- Secrets in a vaultNever in prompts or logs
- Data regionEU Frankfurt
- Support agentRead order #10482 for Priya ShahAllowed
- Support agentRefund of $184 requestedNeeds approval
- Sales agentTried to open the HR folderBlocked
- Ops agentBooked INV-88213 in QuickBooksAllowed
- OwnerApproved the refund in SlackApproved
| Agent | May read | May do alone | Needs a person |
|---|---|---|---|
| Support | Orders, help center | Answer, book, track | Refunds |
| Sales | CRM, website | Reply, qualify, book | Discounts |
| Operations | Invoices, orders | Read, match, draft | Booking above $5,000 |
| Company brain | Per role of the person asking | Answer with sources | Nothing |
Where your data goes, and where it does not.
Four steps for every task. Nothing is stored outside your systems, and nothing is used to train public models.
- Step 1
Your tools
Your data stays where it is: inbox, CRM, drives and accounting.
- Step 2
Your region
The platform runs in the region you choose, with secrets in a vault.
- Step 3
The model
Only what a task needs is sent, over an encrypted business API.
- Step 4
The result
Back in your tools, logged, and approved by a person where needed.
Security, in plain words.
Where is our data stored?
In the region you choose, in accounts that belong to you. We do not keep copies on our side.
Is our data used to train AI models?
No. We use models through business APIs that do not train on what you send, and we only send what a task needs.
Can an agent do something we did not allow?
No. Every agent has an allowlist of actions and permissions per role. Anything outside it is blocked and logged for review.
Can we see what the agents did?
Yes. Every action is in the audit trail and visible in Mission Control, with who approved what and when.