Pricing built around how enterprises actually buy AI agents
Three pricing models, transparent line items, and the controls your finance and procurement teams need — data residency, VPC isolation, audit trails, and named SLAs.
Pricing models
Pick the model that matches your rollout
Most teams start on per-request, graduate to subscription once usage stabilises, and negotiate flat-fee agreements for multi-agent rollouts.
Per request
Pay for what runs — every request, metered.
Best for uneven workloads, proofs of concept, and teams just getting started.
Every agent invocation is metered through the Polsia runtime. You see live usage in the dashboard and pay only for what executed — including failed runs that cost you tokens or tool calls. There is no minimum spend, no seat license, and no commit. Move to subscription once usage stabilises.
What you get
- Per-request unit price scales with agent complexity (tool count + integrations).
- Overage billed item-by-item; no surprise tier bumps.
- Inclusion of API usage, cloud execution, and observability in the request price.
- Ideal for pilots and uneven workloads.
What affects cost
Six factors drive the number on your invoice
We surface every driver in the dashboard, so finance sees real cost — not a black-box estimate.
Tools, integrations, and branching logic behind each agent determine its runtime cost. A single-tool agent and a ten-tool workflow are not billed the same.
Monthly request volume is the headline driver for per-request billing. Most teams stabilise usage within a few weeks and move to a subscription.
Each editor — author, reviewer, ops — is a seat. Read-only auditor access is free across all tiers.
Shared infrastructure is included. Dedicated tenants and BYOC deployments in your own AWS, GCP, or Azure account are priced separately.
Audit logs, runtime guardrails, and policy enforcement are standard on Growth and Enterprise. Continuous audit streaming is an Enterprise feature.
Region pinning (EU, US, APAC) and BYOC residency are part of Enterprise. Multi-region fall-over is available on a flat-fee engagement.
Tiers
Starter, Growth, or Enterprise
Same agent catalogue, three levels of control and isolation — so engineering, security, and procurement can each sign off.
Starter
Free to start · per-request overage
- Pricing model
- Per request
- Included requests
- Up to 1,000 / month
- Editor seats
- 1 seat included
- Isolation level
- Shared multi-tenant
- Guardrails & audit logs
- Basic execution logs
- Data residency
- US (Virginia)
- Support SLA
- Email · 1 business day
Growth
From $2,400 / month · annual available
- Pricing model
- Subscription or flat fee
- Included requests
- Negotiated pool (typically 100k+)
- Editor seats
- Up to 25 seats
- Isolation level
- Dedicated tenant in Polsia cloud
- Guardrails & audit logs
- Guardrails + 90-day audit retention
- Data residency
- US, EU, or APAC pinning
- Support SLA
- 4-hour first response · named CS
Enterprise
Custom — annual flat fee
- Pricing model
- Flat fee (annual)
- Included requests
- Negotiated
- Editor seats
- Unlimited
- Isolation level
- VPC / BYOC in your own AWS, GCP, Azure
- Guardrails & audit logs
- Custom policies + continuous audit streaming
- Data residency
- Region pinning + BYOC residency
- Support SLA
- 1-hour SLA · 24/7 on-call · TAM
FAQ
Common questions from enterprise buyers
Ready for a personalised quote?
Tell us your team size, expected volume, and residency requirements — we'll come back with a tailored breakdown.