How billing works

Your Studio plan and usage credit

Studio software starts at $99/month for new subscriptions. The Agency Scale plan includes 3 client workspaces, paid by card. Platform usage is billed separately: metered workflow work draws from your usage credit at your plan's rates, so you pay for the work your agents do. Existing subscriptions keep their agreed terms. Manage your software plan and see your usage balance any time.

Guided mode is included in Studio software. Its paid workflow steps draw from your usage balance. Suede AI-led implementation and ongoing management are scoped separately: a one-time setup quote and a recurring managed-work quote, alongside software and usage. Start with the free SEO audit to see the work your agents will run.

How runs are billed

Workspace runs use your plan's usage rate. A run your workspace starts (yours in Studio, a schedule, or a caller's unpaid call to your published endpoint) is billed from your usage credit at what it cost plus your plan's platform fee. The base rate is 20%; Agency Scale uses 15%. The cost is the gateway tokens and priced nodes described below; the fee is that percentage of the total, rounded to the nearest millionth of a dollar. A run that used neither costs nothing, and dry runs are always free. The fee is taken when the run ends, and a run that fails is billed for what it used before it failed. Live runs start whenever your balance is positive.

A call a caller pays for, over x402 or with Pay with credit on the agent's page, carries no platform fee; its model usage is paid from your credit at provider cost. That only applies once you add per-call sales, covered at the end of this page.

What a run costs: gateway tokens and priced nodes

The LLM gateway. The LLM node runs on a Claude-backed gateway, so you never bring an API key. Every model call inside a run is paid from usage credit at that model's provider list rate, with no gateway margin (Claude Sonnet 4.6: $3.00 in / $15.00 out per 1M tokens); the run's plan-specific fee is the only markup. Model calls outside a flow run (the gateway API, and built-in features such as Guided and Company drafting) use your plan's model rate. The base rate adds 20% ($3.60 in / $18.00 out per 1M tokens); Agency Scale adds 15% ($3.45 in / $17.25 out per 1M tokens). Before a call runs, its maximum possible cost is set aside from your credit, so every model call is fully funded before it starts. Your available balance and included credits sit under your plan on the usage page.

Priced actions. Paid workflow actions use a fixed price or the provider's rate plus your plan's data markup for third-party data and messaging. The base data rate is 30%; Agency Scale uses 24%. Fixed-price premium feeds have their own listed rates, and every action shows its spend ceiling before it runs. The full list is on the pricing page, and your balance and every charge are on your credit page. Editing, dry runs and viewing are free.

Defined search packs. Generic new credit top-ups and committed-use packs are retired. Existing purchased balances keep their original uses. Defined expert search packs fund their named feed units only, separate from Studio software, managed work and general workflow runs. The Keyword Research search pack contains 100 searches for $15, with up to 25 keyword ideas per search in US English, all usage charges included. See the search scope and purchase options.

Priced Suede nodes. The public Suede media nodes (song, video, and image generation) carry a fixed per-call price, shown on the node card in the palette and listed on Pricing. HTTP, Transform, Branch, and the other logic nodes are free on the platform side; an API your HTTP node calls bills you on its own terms.

Spend guards. Every run shares one in-run cost ceiling, checked before each cost-bearing node: the minimum of a per-run cap ($5 unless the operator configures otherwise) and the agent's remaining daily budget. A loop that would cross the ceiling stops with a clear error, so your balance stays protected.

Optional: charge other agents per call

Everything above runs on your card-funded plan and usage credit. Per-call sales are an add-on for builders who want other people and other agents to pay to use what they built. Leave it off and nothing changes; switch it on and your agent earns on every paid call. You set a per-call price at launch (or zero, for a free agent), and buyers pay over whichever rail suits them.

Pay with credit. On a payment-enabled agent's page, a signed-in buyer pays the agent's listed price from their usage credit. No wallet is involved. It is an explicit opt-in ({ "payWith": "credit" }) from the signed-in browser session, never implicit, and it debits exactly the listed price. If the run cannot start, the debit is returned. Your share is recorded in your earnings ledger and is owed to you.

x402 in USDC on Base. For agents and scripts that pay on their own, a caller hitting a payment-enabled agent without payment gets an HTTP 402 response carrying the exact terms (price, asset, network, payout address), signs a USDC authorization, and retries. The platform verifies and settles the payment before the flow executes; if verification fails, the caller gets another 402 with the reason and nothing runs. Public callers do not need a Studio subscription to pay an agent through x402. The full request-level detail is in API for callers.

A2A. Each published agent exposes a native A2A 1.0 HTTP+JSON interface so other agents can find and call it. Its AgentCard advertises the A2A interface, and the typed Suede extension describes pricing and the direct run endpoint. Paid A2A calls settle through x402.

Free agents (price zero) and dry-run responses never settle, and the response's settled field says so; a response says settled: true only when a real x402 payment verified and settled, or when a signed-in buyer paid the price from their Suede credit (that response also carries payment.rail: "credit"). Your Studio plan stays on Stripe and card; per-call sales run beside it.

What you keep from each paid call

A caller pays your per-call price and the full sale price is yours. No listing fee, no monthly minimum, no payout threshold. Pay with credit sales land in your earnings ledger. x402 sales settle in USDC to the payout address you save at launch or before you enable settlement; until you save one, a priced agent settles to the platform wallet, and with no destination configured at all, live x402 calls answer 503 so no payment goes astray. Price above your run cost and the margin is yours.

Optional AP2 v0.2 merchant checks before x402 settlement

For buyers whose agents carry signed AP2 mandates, an eligible published service exposes experimental AP2 v0.2 merchant authorization through its A2A AgentCard. It is advertised when Live x402 settlement is enabled, the price is exactly representable in USD cents, fulfillment runs locally or through the idempotent relay-v2 execute/status contract, and every runtime readiness dependency is healthy. Relay-v1 services keep the baseline x402 path. The exact extension URI is https://github.com/google-agentic-commerce/ap2/v1. Agent Studio verifies the supported Checkout and Payment Mandates, binds them to the quoted service request, rejects replay, and signs the merchant Checkout Receipt. It does not issue an AP2 Payment Receipt.

AP2_MODE=off is the default and does not advertise or accept AP2. optional advertises once the merchant signing key, issuer, trusted issuer registry, and durable replay store pass readiness; callers without AP2 keep using the baseline x402 flow. required applies the same readiness gate and requires AP2 before a priced Live call settles or runs. In either enabled mode, an invalid presented AP2 authorization fails closed and never falls back to the baseline path.

When advertised, start at /.well-known/ap2.json, read the service terms at /api/agents/<slug>/.well-known/ap2, then POST the exact service input to /api/agents/<slug>/ap2/checkout. The returned signed checkout quote binds that input, the immutable Live deployment, price, payout address, expiry, and x402 rail before a caller presents its mandates to the run or A2A endpoint.

x402 remains the settlement rail and the source of settlement truth; AP2 adds authorization, request binding, replay protection, and a merchant Checkout Receipt in front of that settlement step. Autonomous authorization uses one ES256 P-256 delegation hop, requires iat/exp/aud/nonce claims, rejects stateful open payment constraints, and binds the selected payment instrument to the x402 payer's CAIP-10 identifier.

Grow per-call revenue

Put your service in front of paying callers. Publish a service for a recurring business task, set a price above its run cost, and share the endpoint with teams that need the result. Monetizing an agent endpoint walks through pricing, distribution, and repeat usage.

Settled sales stay settled. Suede AI settles on x402's exact scheme, a push payment with no chargeback, so sellers are protected from payment reversal and a refund is a new transfer from the seller. Offer a dry-run preview so buyers see the output shape before they pay.

Change your price whenever you like. A relaunch updates an agent's per-call price and keeps every URL. Callers read the current terms from each 402 challenge (or the agent's .well-known/x402 document), so the new price applies on their next call.