QUASIHUMAN / v0.4
Pricing

Two ways to run it. Both of them a single-tenant AWS account.

Either you own the account and bring your own keys, or we provision and run one for you. The platform fee is flat and the same product ships in both cases — what changes is who holds the cloud contract, the provider accounts and the bill at the end of the month. Nothing here is priced per seat.

01 / PLANSTwo deployment models

One flat platform fee. The difference is whose AWS account it runs in.

YOUR ACCOUNT

Customer-deployed

$1,499 / month
+ 5% of metered model spend

We deploy the stack into an AWS account you own and operate the fleet inside it. Your cloud bill comes from AWS. Your model keys, and the key for anything else the agents reach, stay in your secrets manager and are billed to you by each provider directly.

Included
Platform licence and fleet operation
Landing zone deployed into your account
Connector library, maintained and versioned
Evals, supervision and audit provenance
A direct line to the engineers building it
You bring
An AWS account and a capped, federated role
Your own LLM provider accounts and keys
Keys for anything else you connect — Firecrawl, search, whatever it is
The infrastructure bill, itemised by AWS rather than by us

The only fee we add on top is 5% of metered model spend. Nothing else you bring carries a markup.

OUR ACCOUNT

QuasiHuman-managed

$4,999 / month
+ 10% processing on pass-through usage

We provision and run a dedicated AWS account on your behalf — single tenant, same architecture, same isolation, just not your paperwork. Usage is billed on at cost with a 10% processing fee, and you get one invoice instead of a provider relationship per vendor.

Included
Everything in customer-deployed
AWS account provisioning and ongoing management
Model and third-party provider accounts held by us
Usage passed through, itemised, on one invoice
One supplier and one purchase order, not four
You bring
One operational workflow that genuinely hurts
Access to the people who actually do that work
Nothing else — no cloud account, no LLM or tool-provider contracts to hold

AWS, model spend and third-party services are all billed at cost plus the 10% processing fee.

Not in either fee

Custom routines — deciding with you which work an agent should take on, then designing the routines and the operating model around them — is a separate engagement, scoped per deployment. It is optional: the colleague learns your work on its own either way.

What that involves →
02 / USAGEWhat carries a fee

Usage is usage. We add a fee to the part we carry, and to nothing else.

Customer-deployed
AWS infrastructure billed to you by AWS
Model spend your provider · + 5%
Third-party services your key · no fee
Seats, conversations, agents not metered
QuasiHuman-managed
AWS infrastructure at cost · + 10%
Model spend at cost · + 10%
Third-party services at cost · + 10%
Seats, conversations, agents not metered

In the customer-deployed model you hold the accounts, so the only thing we can meter is what the agents spend on tokens — and the 5% sits on that alone. Connect Firecrawl, a search API or anything else with your own key and it costs you exactly what that provider charges. In the managed model we hold the contracts and front the spend, so the 10% sits on everything that passes through us, cloud bill included.

Illustrative only · not a quote A month with $900 of AWS, $2,000 of model spend and $200 of third-party services, run through both models.
Customer-deployed
Platform fee $1,499 invoiced by us
Model spend $2,000 paid to your provider
Model fee · 5% $100 invoiced by us
AWS infrastructure $900 paid to AWS
Third-party services $200 paid to each provider
Invoiced by us $1,599
Total outlay $4,699
QuasiHuman-managed
Platform fee $4,999 invoiced by us
AWS infrastructure · + 10% $990 passed through
Model spend · + 10% $2,200 passed through
Third-party services · + 10% $220 passed through
Invoiced by us $8,409
Total outlay $8,409

The gap between the two is what it costs to not have an AWS account, four provider relationships and a cloud bill to reconcile at month end.

03 / INCLUDEDThe same product either way

The architecture doesn't change. Only whose name is on the AWS account.

Single-tenant by construction

One customer is one AWS account — not one row with a tenant_id on it. That holds whether the account has your name on the invoice or ours.

The whole stack, deployed

Agent runtime, durable workflows, pgvector memory and realtime transport, deployed inside the boundary and operated as a fleet.

Connectors, maintained

The maintained library for common enterprise systems, versioned like the rest of the fleet. Bespoke connectors for the software only you run are scoped per deployment.

Supervision and audit

Approval gates on blast-radius mutations, per-run token, cost and wall-clock ceilings with a kill switch, and every action logged, replayable and reversible.

Evals as a merge gate

Self-review on every turn and golden conversations on every release. A change that breaks how an agent handles a real scenario doesn't ship.

Spend you can see

Model spend is metered per run and shown against the work that incurred it, so the usage line is attributable rather than a lump at month end.

NOT PER SEAT

Seats are how we handle identity and approval, not how we bill you. Putting the whole team behind one colleague costs the same as putting one person there.

NOT PER CONVERSATION

A busy month moves the usage line and nothing else. The platform fee is flat.

NOT PER AGENT

The fee is the deployment, not the head count inside it.

The number that matters is the workflow it takes off your desk.

Bring us one operational bottleneck that genuinely hurts and we will scope it properly — which deployment model fits, what it costs to run, and what it is worth before anyone signs anything.

Join the alpha cohort Read the deployment model