Pricing

Priced around your deployment.

AvertGate is deployed through a guided onboarding, not a self-serve signup. Talk to us and we'll scope a plan around your agent traffic.

Proof of Concept

Step 1
2–4 weeks

Run AvertGate against your real agent traffic before committing to anything.

  • Guided setup with our team
  • Your own agents, your own policy
  • Full audit log of what was caught
Book a demo

No public list price and no self-serve signup — every deployment starts with a conversation.

Included

Every deployment gets the whole gate.

There are no feature tiers. What changes between customers is scale, not protection.

Inline MCP proxy

Runs in your network, next to your agents. Tool payloads never leave it.

YAML risk-tier policy

Read passes, writes get argument checks, irreversible actions need a human, unlisted tools are denied.

Two-phase inspection

The request is checked before it runs; the result is scanned for injection before the agent reads it.

Human approval in Slack

Money movement, outbound email and deletion wait for a person. No answer means no.

Hash-chained audit log

Every decision with its reason and rule id, append-only and tamper-evident.

Private, invite-only tenant

Your dashboard, policy editor and agent registry. Access only by invitation.

How buying works

From first call to your own tenant.

  1. 01Demo

    See a real attack stopped

    The same indirect injection, run with the gate off and then on.

  2. 022–4 weeks

    Proof of concept

    AvertGate in front of your own agents and tools, with a policy we write together.

  3. 03Annual

    Contract

    Scoped to your agent count and traffic volume.

  4. 04Go live

    Your private tenant

    Your CTO or IT lead gets a one-time invite link and sets their own password.

FAQ

Buying questions.

Is there a free or self-serve plan?

No. AvertGate governs real credentials, so every deployment starts with a conversation and a proof of concept on your own agents.

What is the price based on?

Your agent count and traffic volume. There is no public list price; we scope it with you after the proof of concept.

What do we need for the proof of concept?

At least one agent that calls tools over MCP, and someone who can point it at the gate. We write the first policy together.

Can it run without any cloud dependency?

The gate itself runs in your network. For air-gapped setups the AI guard can use a local Ollama or vLLM model.

Start with a proof of concept.

Two to four weeks on your own agents. You see exactly what the gate would have stopped.