x402: the pay-per-call pilot

x402 exists in AI Tasker, but not the way it is often described: it is not general pay-per-call billing and it does not replace the wallet. It is a closed pilot, with a single tool, a single test network and a fixed price, designed to prove the pattern without touching real billing. This page states the exact scope and how to check whether it is active in your deployment.

The exact scope, no decoration

Pilot definition in code/mcp-server/src/x402.ts.
Parameter	Value	Why it matters
Paid tool	`get_service_info_paid` (only one)	The rest of the server stays free; there is no paid catalogue
Network	`eip155:84532` (Base Sepolia)	A testnet: no real money is at stake
Scheme	`exact`	One exact amount per call, no auctions or budgets
Price	0.01 USDC	Fixed in code; the client does not negotiate it
Facilitator	Public testnet `https://x402.org/facilitator`	Verifies the payment; if it fails, the call does not run
Quota effect	None (`X402_MCP_PAID_TOOL` = 0 units)	Payment is on-chain; the AI Tasker row is audit only

The pilot sits behind a flag: with X402_ENABLED unset or false, the pilot config returns null and the tool is not even registered. That means the answer to "can I pay per call?" depends on the deployment, and it is checked, not assumed.

How a call behaves

The design is fail-closed, and that is its most important property: the tool handler does not run unless the payment is verified. There is no window where the server does the work and charges afterwards.

The client calls get_service_info_paid like any other tool.
Without a valid payment the call produces no result: the x402 wrapper requires verified payment requirements before entering the handler.
Payment settles on the network with the exact scheme, verified by the testnet facilitator.
After settlement, the MCP server records the operation in the backend with POST /api/x402/settlements, authenticated with the agent’s atkr_ API key.
The row created is X402_MCP_PAID_TOOL with units: 0: it shows up in the audit trail and consumes no workspace quota.

If the facilitator is slow or errors, verification fails and there is neither execution nor record. A visible failure beats a phantom charge: the pilot is built so no payment is left untraced and no work is left unpaid.

How to check whether it is active

Three checks, from cheapest to most conclusive:

The public catalogue (40 tools) does not include get_service_info_paid: it is generated from code and only lists always-registered tools. If it appears there, something changed and should be reviewed.
tools/list on the deployed server is the source of truth: if the tool is there, the pilot is active in that deployment; if not, it is off.
Price and network are read from the tool announcement itself in tools/list, not from this page: if the pilot changes, it changes there.

Looking for the paid tool in the deployed catalogue (needs a valid API key and an MCP session).

curl -s -X POST https://mcp.ai-tasker.theboomer.dev/mcp \
  -H "Authorization: Bearer atkr_your_key" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -H "mcp-session-id: $SID" \
  -d '{"jsonrpc":"2.0","id":9,"method":"tools/list","params":{}}' \
  | jq -r '.result.tools[].name' | grep -i paid || echo "x402 pilot not exposed"

Why it is not the billing model

Three concrete reasons, worth having at hand when someone asks:

It covers one tool, not the product. The other 40 stay available under the workspace quota.
It runs on testnet. Base Sepolia is a test network: it validates the flow, it does not take revenue.
It does not touch the quota. An x402 call spends no credits and does not deduct from the monthly allowance: if it counted as an operation, the pilot would charge twice.

The real model is unchanged: 2.00 € per month per billable unit plus prepaid credits for consumption. Whoever wants that model has the wallet and per-agent budget guides.

What the design protects

Risk	How it is cut
Private keys on the server	There are none: only the public `payTo` address. The process never signs payments
Sensitive data in logs	No payment payloads or signatures are logged; only verifiable identifiers
Double charge on retry	`Operation.idempotencyKey` (SHA-256 of the payment) with a unique index: a replay returns the existing row
Charge without work or work without charge	Fail-closed: no verification, no handler; the record happens only after settlement
Process dying mid-way	The operation stays `pending` and reconciliation reverts it: never a phantom charge
Replay of a reverted payment	`409` conflict, never a silent success

The recording endpoint is deliberately narrow: POST /api/x402/settlements requires API-key identity (403 X402_API_KEY_REQUIRED without it), validates network (X402_UNSUPPORTED_NETWORK), tool (X402_UNSUPPORTED_TOOL) and scheme, and accepts no free-form fields. It is not a payments API: it is the pilot’s audit ledger.

If you build on x402

Check with tools/list whether the paid tool is exposed in the deployment you are going to use.
Treat payment as a precondition, not a result: no settlement, no data.
Keep the paymentId returned by the flow so you can match the payment with the server audit.
Do not use the pilot as a commercial "pay-per-use" argument: real pay-per-use is credits.

Check and use the AI Tasker x402 pilot

Verify whether the paid tool is exposed, understand the fail-closed flow and avoid confusing the pilot with the billing model.

Open an MCP session
initialize and mcp-session-id as with any MCP client; a valid atkr_ API key is required.
List the tools
tools/list and look for get_service_info_paid. If it is absent, the pilot is off in that deployment.
Read the payment announcement
Price (0.01 USDC), network (eip155:84532) and scheme (exact) come from the tool announcement itself.
Call only with verified payment
No settlement, no execution: the wrapper requires payment before entering the handler.
Check the record
After settlement there is an X402_MCP_PAID_TOOL row with 0 units, visible as audit and consuming no quota.

Frequently asked questions

Can I pay for any call with x402?

No. The pilot covers a single tool, get_service_info_paid, on Base Sepolia at 0.01 USDC. Every other operation is paid with the workspace quota and its credits.

Does x402 consume credits?

No. The operation is recorded as X402_MCP_PAID_TOOL with 0 units: it is audit. Payment happens on-chain and never touches the monthly allowance or the wallet.

How do I know it is active in my deployment?

With tools/list: if get_service_info_paid appears, the pilot is exposed; if not, it is off. The public 40-tool catalogue does not include it because it only lists always-registered tools.

What happens if the facilitator fails?

Verification does not pass, the handler does not run and nothing is recorded. The design is fail-closed: a visible failure is better than a charge without work.

Can the same payment be charged twice?

No. The payment identifier is hashed and uniquely indexed: a replay returns the existing row, and if the entry was reverted it answers 409.

Is it real money?

No. The pilot network is Base Sepolia, a public testnet. It validates the flow; it does not bill anyone.

Sources

Public tool catalogue (40, without the paid one)
https://ai-tasker.theboomer.dev/llms-full.txt
Plans, operations and their cost
https://api.ai-tasker.theboomer.dev/api/plans
API OpenAPI contract
https://api.ai-tasker.theboomer.dev/api/openapi.json

Figures verified against the live API and each product’s official page on 21 October 2026. Third-party prices can change: check the official page before you commit.

Related guides

All guides — How to choose a task manager, MCP and agent lifecycle tooling for AI agents.
Credit wallet: top up and control agent spend — How the AI Tasker credit wallet works: which operation consumes a credit, how to top up with packs or AutoRefill, and what happens when the balance hits zero.
Budget and quotas per agent — How to split credits and quotas across several workspaces and agents in AI Tasker: what each wall returns, how to read usage and how to cut spend fast.
Use the AI Tasker API without an SDK — The AI Tasker HTTP API without a library: x-api-key authentication, idempotent task creation, cursor pagination and what to do with each error class.
How agent usage is billed — How AI Tasker measures and bills agent usage: the 2 € billable unit, operations, credit packs and auto-refill. No surprises.
Pricing — Real prices: free plan and a 2.00 €/month billable unit.
MCP integration — https://mcp.ai-tasker.theboomer.dev/mcp
/llms.txt — Machine-readable contract.