Question 1

What is Superagent?

Accepted Answer

Superagent is a three-part system that makes your AI safe and gives you the evidence to prove it. Guardrails are embedded models that prevent failures at runtime. Continuous tests measure whether your system behaves safely. The Safety Page shows controls and test results to customers and prospects.

Question 2

How is this different from generic AI evals or benchmarks?

Accepted Answer

Most evals focus on model performance or generic agent behavior like tool calling. Superagent is focused on AI-system failures, not generic security paperwork. We combine defense, measurement, and proof in a single platform that's usable from day one.

Question 3

What do the guardrails do?

Accepted Answer

Guardrails are small language models embedded into your agents that provide real-time protection. Guard stops prompt injections, jailbreaks, unsafe instructions, and unintended actions. Redact prevents leaks of secrets, internal data, and sensitive customer information. Verify checks outputs for correctness, reliability, and adherence to expected intent.

Question 4

What is the Safety Page?

Accepted Answer

The Safety Page is a customer-facing page that shows your controls, test results, and verification status. It works from day one with pending verification states, so you can use it directly in sales cycles to demonstrate your safety and compliance posture to prospects.

Question 5

Who is Superagent for?

Accepted Answer

Teams building or deploying AI agents who sell to enterprises or regulated industries, need defensible proof of safety for procurement, want to prevent high-impact failures before production, and need something they can show prospects immediately.

Question 6

How do I get started?

Accepted Answer

You can start using Superagent today. Sign up for the dashboard to access the guardrail models and set up your Safety Page. The continuous tests can run immediately against your system. For a walkthrough of how Superagent fits your specific use case, book a call with our team.

Question 7

How much does it cost?

Accepted Answer

Guardrail models are billed on a usage basis according to token consumption. The Safety Page and continuous tests are available as part of our platform plans. The easiest way to get accurate pricing for your use case is to book a call.

Model	Input tokens	Output tokens
Guard	$0.9 per million	$1.9 per million
Verify	$0.9 per million	$1.9 per million
Redact	$0.9 per million	$1.9 per million

Make your AI safe.
And prove it.

Stop failures where they start—inside your agents

Find the gaps before attackers do

Close enterprise deals
without the safety objection

Three parts that work together

Guardrails

Adversarial Tests

Safety Page

Simple, usage-based pricing

Frequently Asked Questions

Lamb-Bench: See how
your model stacks up