Aardvark is free to use forever — prebuilt binaries for macOS, Linux, and Windows, self-host anywhere. Paid plans add managed hosting, seats, support, and a monthly included-AI allowance metered in real dollars — never opaque credits.

Aardvark is free to use, forever: prebuilt binaries for macOS, Linux, and Windows, self-hosting anywhere at no cost, and an open contribution mirror for these docs. Paid plans are for teams that want a predictable monthly bill — managed hosting, more seats, real support, and a monthly included-AI allowance in dollars, keeping pricing simple and dollar-based. Your Aardvark key also grants you access to hundreds of AI models you can use strategically to control costs as you go.

Every model’s real cost per answer, on every plan: the pricing table.

Compare plans

Nothing core is paywalled: self-hosting, Markdown-for-Agents, llms.txt, the agent-skills index, whole-site PDF, and every component and theme ship on Free. Paid plans add hosting, seats, support, controls — and cheaper AI.

The honest AI meter

Every competitor meters AI in credits you can’t reason about. Aardvark bills AI in dollars: the pricing table publishes what every model costs per answer on every plan — real figures, no conversion math. Subscribers pay less per metered dollar the more they commit (≈7% off on Pro, ≈13% on Business, 20% on Enterprise), on top of the allowance their plan includes. All of it runs on Aardvark’s managed keys — you never create, rotate, or secure a model-provider key. Enterprise accounts can optionally register their own provider key so upstream calls run on their own provider contract; the metered price is the same either way, so it is a contracting choice, not a discount.

Your included allowance is denominated in those same billed dollars:

• The dashboard shows dollars included, dollars used, your own trailing burn rate, and an estimated depletion date — no “≈N answers” marketing math.
• Unused allowance rolls over one month, up to 2× your monthly amount.
• Answers served by :free models never bill and never touch your allowance.

When the included AI runs out

Cap-and-hold is the default. Paid AI pauses at your cap and you’re notified, so it cannot add further overflow charges. Your “max possible bill” is the subscription price plus any fixed add-on-seat charges and any overflow already incurred this month. (Usage-based GitHub Automations compute, if you use it, draws your prepaid balance separately and isn’t included in that figure.) Prefer to stay unstuck? Two self-serve options, both under your control:

Annual billing

Annual plans take 20% off the platform fee — the included-AI allowance is funded monthly at full value either way (it’s real usage, not a discountable line item), so the effective saving is about 12–13% off the yearly total — ~12% on Pro, ~13% on Business (the bigger a plan’s fee is relative to its bundled allowance, the more of the bill the discount reaches). Pro: $1,046.40/yr (vs. $1,188 at par, ~12% off). Business: $3,638.40/yr (vs. $4,188, ~13% off). The allowance still resets every month on annual plans.

FAQ

Billing is designed to be predictable and never to surprise you. Here’s exactly what happens in the situations people ask about most.

Every question — and its exact answer — now lives in the searchable support knowledge base, organized by category with the most-asked questions up top.