Pricing

The controls are free. The savings pay for the license.

Budgets, pools and hard gateway enforcement cost nothing — forever, for unlimited people. The paid ROI engine recovers idle credits and cuts token waste, so it funds itself.

Validate before you buy: a free 60-day ROI Pilot, activated from your console when your real data is ready · no credit card · self-hosted, your data never leaves

Core

$0 forever · unlimited people

Everything you need to give AI budgets real teeth.

  • Credit management
  • Credit pools — company → department → team → project
  • Allocation down the real org tree + peer transfers
  • Cascading budgets — company → department → team → person, run on demand
  • Seats & licenses tracking, waitlists, vacation mode
  • Dashboards + a full append-only audit log, searchable in the console
  • Role-scoped privacy (admin / finance / manager / lead / self)
  • AI Gateway — included free
  • Managed per-user API keys (mint / replace / revoke)
  • Hard budget enforcement — requests stop at ◈0
  • Real-time per-user attribution across providers
  • One-click employee machine setup (.bat / .sh — Windows / Mac / Linux)
  • Per-provider key scoping — a key only reaches the providers its holder has a seat on
  • Deployment
  • Self-hosted (Docker / Kubernetes), multi-provider
pays for itself

ROI Engine

from $6,000 / year — one flat license, banded by the AI spend it governs

Everything in Core, plus the layer that recovers money.

  • Token savings
  • Response cache — an identical repeat prompt costs 0 tokens
  • Smart model routing — simple prompts are served by the cheaper model
  • Measured savings, in dollars, on the Savings & ROI screen
  • Automation
  • Idle-credit sweep — recover unused credits back to the pool, run on demand
  • Application (service-account) pools — apps & bots get their own API key and enforced budget, with recurring funding schedules you run on demand
  • Financial truth
  • Reconciliation against real provider spend
  • FinOps chargeback / showback per team & department
  • Insights — who's starving, who's hoarding, what to fix
  • Support
  • Priority support · SSO/SAML assistance

One price — a flat annual license, banded by your AI spend

Self-attested at purchase from your provider invoices, trued-up once a year. No meters, no per-seat counting, never a percentage of your spend or "savings". This table is the whole price list.

Annual AI spend under managementROI Engine license
Under $125K$6,000 / year
$125K – $250K$12,000 / year
$250K – $1M$24,000 / year
$1M – $3M$48,000 / year
Over $3M$96,000 / year

The 3× rule: you buy only after your 60-day ROI Pilot's CFO-ready memo shows at least 3× the license fee in annualized identified value. If it doesn't — stay free, or extend the pilot.

Spending under ~$60K/year on AI? The free tier is probably the right product for now — your pilot memo will tell you.

Every band includes every ROI Engine feature, unlimited people. Annual billing. Renewal is a new expiry on the same license key — nothing reinstalls.

The ROI Pilot — prove it before you pay

The pilot isn't a timer that starts while you're still installing. It starts when your data is ready to prove something.

1 · Install free

~10 minutes, self-hosted. The controls and the gateway are free forever — this step alone is worth the install.

2 · Run on real usage

The console computes what the ROI Engine would have done with your actual traffic — idle credits it would sweep, spend it would reconcile.

3 · Activate when ready

Once there's real signal, activate the 60-day full-product pilot from inside the console. One click, no credit card, once per company.

4 · Decide on the memo

The pilot ends with a CFO-ready savings memo on your own numbers. The 3× rule decides — not a salesperson.

Limited — first 10 companies

Founding Partners

Six months of the ROI Engine free, your band price locked for life, and a direct line to the founder — in exchange for a named reference and honest feedback.

Apply — hello@allotmentai.com

What's in each tier

FeatureCore — freeROI Engine
Credit management
Credit pools (company / dept / team / project / vacation)
Allocation down the org tree + peer transfers
Cascading budgets down the org tree (run on demand)
Requests & approvals workflow
Seats & licenses, waitlists, idle-seat signals
Vacation mode — a wallet parks into the team pool, returns on comeback
Balance alerts & reserve thresholds
Dashboards & full append-only audit log
AI Gateway
Managed per-user keys (mint / replace / revoke)
Hard budget enforcement — requests stop at ◈0
Real-time per-user attribution (all providers)
Per-provider key scoping — a key never spends another provider's credits
PII redaction — configure a Presidio sidecar in your own LiteLLM gatewayyou configureyou configure
One-click machine setup (.bat / .sh) · GPO / Intune env-var reference
Application (service-account) pools — per-app API keys, spend billed to the pool, recurring funding schedules (run on demand)
Response cache (0-token repeat answers)
Smart model routing (simple prompts → the cheaper model)
Automation & intelligence
Idle-credit sweep back to pools (run on demand)
Reconciliation vs. real provider spend
FinOps chargeback / showback + Savings & ROI
Insights & recommendations
Deployment
Self-hosted — data never leaves your perimeter
Unlimited users & providers
SupportCommunityPriority

The math that closes the budget meeting

On $1M/yr of AI spend, recovering just 5–10% of idle credits returns $50k–$100k — against a $24,000 license.

That's a 2–4× return before counting cache and routing savings — and you only ever buy after your own pilot memo shows at least 3×. The free tier proves the controls; the ROI Engine pays the bill.

Fair questions

What exactly happens when a license expires?

Nothing breaks. After a 14-day grace period the install degrades to Core: pools, allocation, transfers, seats, dashboards, audit — and the gateway with hard enforcement — keep working. Only the ROI Engine features (cache/routing savings, the idle-credit sweep, reconciliation, FinOps, insights, creating new application pools) switch off until you renew. Existing application keys keep working and keep billing their pools — nothing in production breaks. Same license key, new expiry.

Why are application (service-account) pools paid?

Human budgets are free — every employee gets an enforced budget at no cost. Application pools give apps and bots their own API key, a budget billed to a shared pool, and recurring funding schedules you run on demand — that's production workload governance, the layer organizations adopt once AI is in their products. People free, machines paid.

Why is hard enforcement free? Isn't that the valuable part?

Budgets without teeth are advisory — so enforcement belongs in the free tier. We charge for what recovers money: cache and routing savings, the idle-credit sweep, reconciliation and the FinOps layer. You should never have to pay just to make a budget real.

How do you know which spend band we're in?

You attest it on the order form — a real number straight off your OpenAI/Anthropic invoices, trued-up once a year at renewal. We never see your data (the product runs entirely in your perimeter), so there is no meter and nothing to dispute: your own console shows you the same figure the whole time.

What if our AI spend grows mid-year?

Nothing changes mid-term — your price is fixed for the year. At renewal you true-up to the band you're actually in. Bands are wide (4× steps), so normal growth rarely moves you more than one step.

How does the 60-day ROI Pilot work?

Install free and let it run on real usage. When your console has enough data to prove something, it offers the pilot — one click, full product for 60 days, no credit card, once per company (work email). It ends with a CFO-ready savings memo on your own numbers; the 3× rule decides. If the memo doesn't clear the bar, stay on Core — it's free forever.

Do you see our usage or data?

No. The product runs entirely inside your perimeter. The only thing that ever reaches us is an anonymous monthly license check (license id, version, install id, and one aggregate seat count) — no names, no usage, no prompts. It even works fully air-gapped.

Prove it on your own data, free.

Install in ~10 minutes. The controls are free forever; the ROI Engine gets a 60-day pilot — if it doesn't pay for itself, stay on Core.