A GuardBot scoped to your traffic
We map your applications, agents and tools to a starting policy: phrases that must never leave your business, tools an agent may call, size and rate ceilings, and whether a match blocks or is only recorded.
REQUEST A GUARDBOT
Tell us what your AI is allowed to do. We will build the GuardBot that enforces it.
GuardBotAI is an AI runtime security proxy: every request from your applications and agents passes through it and is checked against your rules before it reaches a model or a tool. This page is for teams who would rather have it scoped with us than configure it themselves.
The triple-layered sandbox, configured
Every guardbot starts in Practice, moves to Leashed with one approved destination and an allow-list, and only reaches Live with a deliberate, recorded approval. Breakout signals drop it a layer instantly and alarm you.
A written quote, not a surprise invoice
Pricing is set against request volume, business units and deployment model, then put in writing with usage allowances, response targets and the exact scope of coverage before anything is signed.
How a request works
- Step 1. You send the form below with what your agents do and roughly how much traffic they carry.
- Step 2. We reply with questions if anything is unclear, then propose a starting policy you can read in plain language.
- Step 3. You get a written quote and, if you want it, a walkthrough of live traffic being blocked in a test workspace.
- Step 4. Terms are agreed, your workspace is provisioned, and your first guardbot goes in at the Practice layer.
Self-serve plans are available immediately on the pricing page. Larger organisations with procurement, security review and contract requirements should start on the enterprise page, which sets out terms and the procurement path.
Request a GuardBot
The more you can tell us about what your agents touch, the more precise the first policy will be. We reply to every enquiry ourselves — there is no automated qualification.
We hold no external security certification today and do not claim one. What we do hold is published on the trust policy page and measured on the status page.