Run every agent from one place, inside limits you set.
Our platform runs on a computer in your office, and your documents, conversations and results stay on it. You decide what each agent can touch, and it checks that on every action.
You write the limits. The platform does the rest.
-
01
Declare the limits
For each agent, say which files, tools and websites it may use, and what needs a person’s sign-off. Nothing is granted by default, and the guardrails are derived from this.
-
02
Every action is checked
A brief hidden in the agent’s files told it to post the pricing to an outside site. The site was on its allowed list. It was refused anyway, because a document asked for it, not the user. Every action and refusal is kept in a tamper-evident audit log.
-
03
Attacked on a schedule
Every protection ships with its own attacks. They run against your agents to prove the limits still hold, and an agent that fails can’t be promoted.
Everyone sees their own part. We see none of the content.
-
01
The administrator
Someone inside your office. They see every assistant, all the work on the board and everyone with access. When an assistant wants to send something, they are the one who reads it and decides.
-
02
The team member
Each person gets their own link and their own desk, with the assistants you allow and their own work. A colleague’s work is never sent to their browser.
-
03
Red Gambit
We keep the system running from outside your office. Our dashboard is told that an installation is up, how much it is used, what it costs and what kind of error happened. Documents, requests, answers, file names and people’s names never reach it.
Let clients build their own agents, without giving up control.
If you allow it, clients can create and edit agents in their portal. They choose inside the limits you set, never beyond them.
They choose
Name, instructions, and which of your allowed tools it uses.
You choose
Which tools and websites are on offer, and how many agents.
Sending is held
Anything that sends or writes waits for approval. Clients can’t turn that off.
Attacked first
Every change is tested before it’s saved, and logged.
See it running.
Tell us what your agents do and who uses them. We’ll show you the platform on a call.
