Pricing

Priced per agent. Never per event.

Billing per event would pay you to record less. So recording everything costs the same as recording something: a flat band by agent count, unlimited actions — and published, which the tools quoting you privately will not do.

Self-serve

Free
while you evaluate

Wire up one agent, check one record. Where developers start.

  • Records every kind of action an agent takes
  • Sealed so an edit cannot be hidden
  • Your auditor checks it offline, always
  • What each agent costs, and a ceiling
  • Community support
Start free

Team

Most common
$1,500
per month, billed annually · up to 10 agents

Sealed, checked, kept. Recording more never costs more.

  • Everything in Self-serve, for up to 10 agents
  • Unlimited actions per agent — recording more never costs more
  • Rules that hold an action until a named person signs
  • A packet built for the examiner: FINRA, EU AI Act, Colorado
  • A live page for security reviewers
  • Cost per agent, per run, per customer
  • Kept for years, not days
  • Countersigned by an independent timestamp service
Start free, then upgrade

Scale

$4,000
per month, billed annually · up to 50 agents

For a fleet, and the security reviews that follow one.

  • Everything in Team, for up to 50 agents
  • Examiner packets on a schedule, not on request
  • Write-once storage and legal hold
  • SSO and audit log export
  • Named support with an SLA
Talk to us

Enterprise

$96,000+
per year · unlimited agents, self-hosted

For regulated deployments that run it themselves and bring their own security review.

  • Everything in Scale, no agent limit
  • Runs entirely inside your own cloud
  • We never see the contents — only fingerprints of them
  • Your own DPA, and a custom retention schedule
  • Named support with an SLA
Talk to us
  • No card to start
  • Nothing is blocked until you say so
  • Your proof stays yours if you leave
What actually happens

A week, and then nothing to do.

No implementation project, no consultants, and nobody on your team has to become a compliance expert.

  1. 01Day one

    Turn it on

    Two lines in one agent, or a config change if your calls already run through a gateway. It records; it blocks nothing.

  2. 02Week one

    Watch what it caught

    Every action shows up as you go. You see what your agents have been doing before you decide to stop any of it.

  3. 03When you are ready

    Draw the lines

    Write the limits in plain English. Each rule starts in monitor mode, so you see what it would have stopped before it stops anything.

  4. 04When they ask

    Hand over the packet

    One file for the examiner, the auditor, or your customer's security reviewer. They check it themselves.

Why the meter matters

A meter is an instruction.

Whatever you charge for is the thing your customer does less of. Here are both meters, on the same day of the same agent.

billed per eventthe incentive
01sample at 10% cheaper — and 9 in 10 actions gone−90%
02drop tool calls cheaper — and the actions are the evidence−40%
03shorten retention cheaper — and the question arrives later−70%
04record one agent only cheapest — and the fleet is unrecorded−95%
The cheapest configuration is the one that records least.
billed per agentthe incentive
01sample at 10% no saving$0
02drop tool calls no saving$0
03shorten retention no saving$0
04chain everything same price$0
The cheapest configuration is complete coverage.
Per agent
flat, by band
Per action
$0
Cap on actions
none

It also settles the question nobody wants to be asked: whether our revenue grows when your agents misbehave more. It does not.

And the invoice has no surprises in it.

Keeping years of records has a real cost, so it is a line item rather than an upsell that arrives with the renewal — the invoice nobody budgeted for, on the account you were about to expand.

Everything on the right is metered by somebody in this category. None of it is metered here.

not billedever
retries the same action, attempted twice
heartbeats how a quiet agent stays distinguishable from a dead one
policy checks every decision, allowed or refused
verification yours, your auditor's, or a stranger's
an agent in staging if it shares the production identity
Questions people actually ask

The ones that decide it.

Including the one most pricing pages avoid: what happens to your evidence when you stop paying for it.

01Why per agent and not per event?
Because per-event billing on an audit product pays you to record less, and an incomplete record is the one failure this category cannot survive. Per agent, recording everything costs nothing extra.
02What counts as an agent?
One deployed agent. Twelve support bots is twelve; the same bot in staging and production is two, unless they share an identity. Retries and internal bookkeeping are never billed, and you move band only when you pass the count — never mid-year.
03Why is this published when Vanta and Drata quote privately?
Because a price you have to ask for is a price that depends on what they think you can pay. Ours is on this page. You can decide whether we are worth it before you talk to us.
04Do you see our prompts?
Not by default. We record what happened and a fingerprint of it — the contents stay in your storage, and the record still proves nothing was altered.
05What happens if we stop paying?
You keep your evidence. Export is always available and it checks out without us. A record you can lose is not a record.
Start

One agent, free, today.

Wire up one real agent and check one real record. That is the whole evaluation.

  • No card to start
  • Nothing is blocked until you say so
  • Your proof stays yours if you leave