vaaya / pricing
Give your agents a balance to spend. Pay only for successful calls.
Vaaya shows your agent the exact price before a call runs, meters it to the cent, and charges only when the call succeeds. There are no subscriptions, no seats, and no per-vendor invoices to reconcile at the end of the month. Start with the spending limit your GitHub track record unlocks, or buy a $10, $30, or $100 pack that never expires.
npx @vaaya/mcp installWhere the balance comes from
Three sources, and your agents spend all of them the same way.
Welcome balance
Every account starts with $2 at signup, free and with no card required. Everything below is added on top of it.
A spending limit from your track record
Connect your GitHub and Vaaya reads your public history: how long you have been contributing, what you have shipped, and how consistently. A mid record unlocks around $20, a strong one around $75, and a top record runs into the hundreds. It is added to the free $2 rather than replacing it.
Prepaid, if you prefer
Buy $10, $30, or $100 outright. At catalog prices $10 covers roughly 1,000 web searches, 200 contact lookups, or 30 generated images. Packs never expire, with no monthly minimum and no seats.
What determines your GitHub-backed limit
Here is exactly what sets the number, how it moves, and what your card is for.
What Vaaya looks at
Public data only, with the read:user and user:email scopes: your contributions, merged pull requests, the substance of your public commits and issues, account age, and how consistently you have been active.
What Vaaya never accesses
No repository scope is ever requested. Vaaya cannot see private repositories, and it never gets write access to anything.
How the limit is set
Those signals produce a score. The score sets your starting limit and the ceiling it can grow toward; brand-new accounts without a merged external pull request start at zero and earn their way up.
How it moves
Clean settled use grows the limit. Idle weeks decay it slowly toward a floor, and a failed settlement drops it several times faster than clean use builds it: build slow, lose fast. It never reaches zero, so an account can always recover by paying for what it uses.
When it is recalculated
The GitHub score refreshes at most once a day, and immediately when the scoring model changes. The limit itself moves per settlement event rather than on a schedule, so it reflects what your account has actually done.
How large it gets
Scores run 300 to 850. Below 500 the starting limit is $0 and you earn up from there. Around 600 it starts near $20, around 700 near $75, and a top record starts in the hundreds. Settled use grows it toward a ceiling that scales with the same score, capped at $5,000. Your exact figure is on your account.
A card backs the part above $2
The first $2 is free and needs no card. Anything your GitHub limit adds on top of it is card-backed: the card is what settles what your agents draw from that limit, so it has to be on file before they can spend it.
Settled on a schedule
Vaaya settles what your agents have drawn against that card at the end of each week, or immediately once the amount passes $50.
How your agent pays for a call
You always see one Vaaya price and one ledger entry. Vaaya chooses the settlement rail behind the scenes. The detail below is for developers who want to know what those rails are.
A payment per request
The service answers an unpaid request with a price. Vaaya pays it inline, in USDC on Base, and the request goes through. Nothing is pre-purchased, and the amount settled is the amount the merchant asked for.
A token per purchase
Stripe’s Machine Payments Protocol works the same way with a card behind it: Vaaya mints a payment token scoped to that one purchase, and the merchant charges against it.
Usage-based accounts
Plenty of providers still sell a normal account billed by the second or by the unit. Vaaya holds that account, tracks what your calls consume, and prices each one for you.
Vendors that only sell subscriptions
Some services have no per-call product at all. Vaaya subscribes with its own card and exposes the service to your agent per call anyway, so you are not buying a seat to make one request.
Why this needed new rails
Buying anything online used to assume a human at a checkout page: an account, a saved card, a form to fill in. An agent working through a task cannot do any of that, so in practice it could only use services someone had already signed up for and wired in by hand.
What agentic commerce means here
A payment can now ride along with a single HTTP request, so software can buy one thing, once, at a known price, and be refused cleanly if it cannot pay. Your agent never holds a card or a vendor key. It asks Vaaya, Vaaya pays the provider, and you get one line in your history with the price and the result.
How Vaaya pricing works
Short answer: the catalog price is the final price. Vaaya’s fee is already inside it, and nothing is added after a call runs.
Where Vaaya holds the account
Most of the catalog. Vaaya pays the provider on its own account and the listed price covers that cost plus Vaaya’s margin. You are never billed the provider’s invoice separately, and you never see a line item you were not quoted.
On the pass-through rails
x402 and MPP settle per request against the merchant directly. You pay exactly what the merchant asks for, with no Vaaya markup on top today. The listed price acts as the ceiling your call cannot exceed.
There is no platform fee, no per-seat charge and no monthly minimum. If Vaaya ever changes how it takes margin, that change will show up in the catalog price you are quoted before a call, which is the only number that ever reaches your ledger.
How a call is charged
The same rules apply to every service in the catalog.
Price shown before every call
Your agent sees the exact price and a spend cap before each call, so nothing runs by surprise.
Failed calls are free
If a call fails, it is not billed. You pay only for successful results.
Usage priced as it runs
Models are metered per token and sandboxes per second, on actual usage, for exactly as long as they run. Your agent sees the price or the spend cap before it starts.
Every call in one history
Each call lands in the same ledger with its price and its result. There are no per-vendor invoices to reconcile.
Representative prices
A sample of per-call prices across categories. Your agent always sees the exact price before it runs; these are typical starting points, not a full list.
LLM answers
300+ models behind one OpenAI-compatible endpoint, metered per token.
Web search
Semantic, keyword and cited-answer search across five engines.
Web scraping
Clean page extraction, whole-site crawls, and the exact fields you name.
Deep research
Multi-step, cited research synthesized across the live web.
Lead enrichment
Verified emails, phone numbers, and full profiles, per contact.
Company signals
Funding rounds, acquisitions, job changes and hiring, as they happen.
Image & media generation
Generate images, audio, and b-roll from the top model providers.
Inboxes your agent owns: send, receive, and act on the replies.
Browser sessions
A real cloud browser for clicking, typing and filling in forms.
Compute & GPU
Sandboxes and GPUs from T4 to H100, billed only while they run.
Property data
US property records, valuations, listings and owner data.
Storage & memory
Persistent files and durable semantic memory your agents recall across runs.
Give your agents a balance to spend.
Sign up and your agents have $2 to spend straight away. Connect GitHub when you want a larger line, and you are charged only for the calls that succeed.
npx @vaaya/mcp installNot a terminal person? Set up with a few clicks in Claude, ChatGPT and more. Or skip setup entirely and chat in the browser, with the same services and the same balance.