Pricing — AGX

Commercial path

Evidence first. Rollout second.

You do not move to the next step unless the data and the math support it. Every stage has a clear output and a go/no-go decision.

Evidence Diagnostic

The automated Go/No-Go readiness report. AGX tests whether your data can support the math before you spend a dollar.

Falsification Pilot

A fixed-scope engagement to shadow-test candidate guardrails against your historical data. AGX validates the guardrail before it touches production.

Production Enforcement

Continuous, inline policy enforcement for live AI agents. You pay for each distinct operational boundary AGX actively governs in your critical path.

Pricing principle No workflow protected, no workflow billed.

AGX does not charge for agent chatter, raw API volume, or observer seats. Pricing follows the operating boundaries AGX enforces and the value those boundaries protect.

Pricing model

Three steps. Each one earns the next.

The diagnostic qualifies the data. The pilot shadow-tests the policy. Production only scales where the value is validated by evidence.

01. Evidence Diagnostic

The automated Go/No-Go readiness report. We test whether your data can support the math before you spend a dollar.

Run the Diagnostic

Run the local extractor against your warehouse
Map one measurable AI decision to your existing logs
Receive the evidence-based decision on whether a pilot is worth funding
Output: definitive Go/No-Go recommendation and data-gap analysis

Proof stage

02. Falsification Pilot

$15,000-$50,000 flat fee

A fixed-scope engagement to shadow-test candidate guardrails against your historical data. We validate the guardrail before it touches production.

Scope a Pilot

4-8 week pilot
Define the exact decision boundary and hard constraints
AGX drafts candidate guardrails and shadow-tests them against historical logs
Reject weak policies, tune the survivors, and measure projected recovered value
Output: validated guardrail and executive evidence package required to fund production

03. Production Enforcement

$60,000-$250,000+ / year

Continuous, inline policy enforcement for live AI agents. You pay for every distinct operational boundary AGX actively governs in your critical path.

Discuss Production Scope

Runtime enforcement: sub-100ms inline interception and blocking of out-of-bounds agent actions
Bounded execution: agents operate strictly within approved policies with explicit fail-open or fail-closed degradation rules
Continuous learning: live outcome tracking adapts, confirms, or revokes guardrails based on actual drift
Audit trail: immutable logging of every intercepted decision, policy override, and agent failure
Pricing model: base subscription tied to workflow scale and complexity, with optional capped upside tied to validated recovery

Commercial guarantee

The AGX Commercial Guarantee

The model is designed so you pay for governed operational boundaries, not arbitrary usage meters.

✓

No per-API call taxes.

We do not charge you for the volume of data your agents process. We do not penalize you for scaling.

✓

No seat fees for auditors.

Compliance, legal, and engineering teams need read-only access to the audit logs. We do not charge per user to look at the math.

✓

Evidence first, rollout second.

You do not move to the next step unless the data supports it. The diagnostic qualifies the data, the pilot shadow-tests the policy, and production only scales where the value is validated by evidence.

FAQ

Buyer questions.

The commercial path is built to reduce buyer risk: start read-only, test before rollout, and scale only after the guardrail is worth enforcing.

What determines if the pilot is $15k or $50k?

Scope and data complexity. Clean, narrow workflows sit at the lower end. Multi-system mapping, complex domain rules, and massive historical datasets move it to the higher end.

Does the diagnostic or pilot change production systems?

No. Both are strictly read-only. They use historical exports, controlled views, or Snowflake BYOC. No live writes, no workflow interruptions.

How is the annual production contract priced?

It combines a base subscription tied to workflow scale, connected systems, and evidence volume. Some contracts include capped upside tied to validated financial recovery.

What if the pilot does not validate value?

Then AGX records exactly why: weak evidence, unsupported causal claims, or unacceptable review burden. That is a successful diagnostic because it prevents a bad guardrail from entering production and wasting your engineering budget.

Pay for enforced boundaries, not API calls.