One ticketing platform. Clear tiers. No per-agent fees.

Per-agent pricing makes teams ration the queue

The per-agent number looks small next to what the queue is worth. The damage is not the number. It is what teams do to avoid paying it.

• The compliance reviewer who approves four payments a month shares a login instead of holding one, so the audit trail names the wrong person.
• The night-shift operator works out of a team account, and the handoff record shows a role rather than a decision-maker.
• Auditors never get read-only access, so evidence leaves the system as a spreadsheet export that nobody can tie back to a ticket.
• A contractor waits a week for a seat, and the tickets they were hired to clear queue up behind a procurement approval.

Each of those is a control failure caused by a pricing model. A system that charges for the reviewer discourages you from having one.

Per-agent pricing is not dishonest, and it scales down well. A two-person team pays for two people and knows exactly what next month costs. It stops making sense at the point where the people who need to see a ticket outnumber the people who work it - reviewers, approvers, auditors, and the second team an escalation policy hands the ticket to.

Band pricing has its own edge to be honest about. A band has a ceiling, and passing it moves you to the next tier rather than to a larger invoice per head. Inside a band, adding a read-only reviewer costs nothing, and that is the behaviour worth encouraging.

What changes when you run it yourself

Self-hosted deployment arrives on the Enterprise tier, and it moves both sides of the ledger. Read this part before you compare monthly numbers.

Cost stops tracking headcount and starts tracking infrastructure

The licence is still not priced per agent. What moves onto your budget is the infrastructure: the servers the queue runs on, the storage the audit trail accumulates, and the inference bill for whichever model you plug in. A team running an open-weights model on hardware it already owns pays nothing per token. A team pointing the same plugin at a commercial API pays that vendor directly, on its own contract, and can switch providers as a configuration change.

Control stops being a clause and starts being a boundary

Self-hosted means the ticket data, the models, and the identity provider stay inside your network. There is no vendor-side copy of the queue to negotiate a processing agreement over, and no model endpoint outside your perimeter reading customer messages. In an air-gapped environment the whole loop runs with no egress: intake, triage, approval, execution, and the record it leaves. The corrections your operators make stay training signal on your side of the line, so the queue sharpens on your tickets without those tickets going anywhere. Data residency becomes a deployment fact rather than a contract clause.

You take on the operational work

Self-hosting moves upgrades, backups, restore testing, and uptime onto your team. That is a real cost, and it is the reason the hosted tiers exist. If nobody on the team wants to own a database at 2am, take the managed option and revisit the question when a customer contract, a regulator, or a data residency rule makes the boundary non-negotiable.