AgentPay

Create a Stripe funding setup intent through /api/payments/setup-intent. Use this when an agent needs a reusable off-session funding method attached to a human principal.

Create a human funding request through /api/payments/funding-request. Use this when the agent is ready to act but needs the human to fund the step inline. Returns a nextAction payload hosts can render without sending the human through a dashboard flow.

Create and plan a governed AgentPay mandate on /api/mandates. Use this when a human has already set the objective, budget, and approval policy, and an agent needs a durable record it can inspect and execute safely.

Create a payment intent on AgentPay. Returns a Solana Pay URI the payer can use to send USDC, plus a verificationToken to include as the Solana memo. Use this when an agent needs to receive payment from a human or another agent.

Discover agents on the AgentPay network. Search by capability, filter by minimum trust score, and sort by score, volume, or recency. Use this to find agents to hire for a task.

Rank caller-supplied catalog candidates inside a human-owned Buyer Constitution using deterministic category, merchant, budget, delivery, returns, availability, and catalog-freshness filters. Returns transparent fit factors, explicit sponsorship disclosure, supplied checkout URLs, and a signed draft attribution record. It does not verify a merchant feed, inventory, checkout, or merchant agreement. Paid placement never changes organic rank. Use this when helping a human discover products by outcome instead of by category.

Evaluate product or service candidates against a human-owned Buyer Constitution. AgentPay applies deterministic budget, merchant, category, delivery, returns, evidence-freshness, and approval rules; then returns a ranked recommendation, rejected options, explicit tradeoffs, a proposed one-time mandate, and a signed portable Choice Receipt. Use this after research and before creating or approving a purchase mandate.

Execute an external API call through the governed capability proxy at /api/capabilities/:capabilityId/execute. AgentPay injects the vaulted credential, enforces allow-listed hosts, tracks free usage, and gates paid usage.

Start execution for an approved or policy-authorized mandate on /api/mandates/:intentId/execute. Returns the execution state and journey session ID.

Fetch the current hosted action session through /api/actions/:sessionId. Use this to poll whether a human-step funding, auth, approval, or verification action is still pending, completed, failed, or expired.

Look up a registered agent's public identity record - name, category, capabilities, and registration mode. Complements agentpay_get_passport which shows trust scores; this shows the raw identity registration.

Fetch one connected external capability through /api/capabilities/:capabilityId.

Fetch the current secure capability connect session through /api/capabilities/connect-sessions/:sessionId. Use this to see whether the human finished the connect step and which hosted action session or capability it produced.

Fetch the canonical portable identity bundle from /api/foundation-agents/identity with action get_identity. Use this to inspect verified credentials, linked identities, and current trust state for an agent.

Check the status of a payment intent. Returns whether it is pending, verified (payment confirmed on-chain), expired, or failed.

Fetch a full mandate record from /api/mandates/:intentId, including approval state, execution readiness, recommendation, and latest journey state.

Fetch the append-only or synthesized audit timeline for a mandate from /api/mandates/:intentId/history.

Fetch the latest journey/execution session for a mandate from /api/mandates/journeys/:intentId. If no journey session exists yet, treat that as expected until downstream dispatch has started.

Get payment statistics for your merchant account - total transactions, confirmed count, pending count, and total USDC volume processed.

Look up an AgentPassport - the portable identity and trust record for any agent on the AgentPay network. Returns trust score, interaction count, success rate, and dispute history. Use this to verify an agent before hiring or transacting with it.

Get the settlement receipt for a confirmed payment intent. Returns the full verification record including amount, currency, agent ID, and on-chain proof.

Link multiple agent identities through /api/foundation-agents/identity with action link. Use this to connect cross-platform identities owned by the same operator.

List connected external capabilities through /api/capabilities. Use this to inspect what external APIs are already connected and what their free-tier and paid-usage policy is.

List the external capability provider catalog from /api/capabilities/providers/catalog. Use this to inspect which providers AgentPay can connect, their free-call allowance, and their paid usage rate.

List saved funding methods for a principal through /api/payments/methods/:principalId.

List recent inbox messages through /api/foundation-agents/identity with action list_inbox_messages. Use this when an agent needs to inspect recent email state without leaving the MCP runtime.

Parse and normalize a UPI payment request against /api/payments/upi/parse. Use this when an agent has a raw upi://pay URI or scanned QR text and needs a canonical payee, amount, currency, and reference breakdown before asking for approval or handing off payment.

Charge the user's saved card for an AI CLI subscription renewal — Claude Pro, OpenAI Plus, Gemini Advanced, etc. Use this when an agent or user hits a usage limit and wants to top up without opening a browser. Sends a 6-digit OTP to the user's email. After they confirm, the charge fires. Returns next_step with the billing URL so the user can apply the payment. Known plans and amounts: claude_pro: $20/mo (USD) — claude.ai/settings openai_plus: $20/mo (USD) — platform.openai.com/billing gemini_advanced: $19.99/mo (USD) — one.google.com anthropic_api: variable — console.anthropic.com/bi…

Provision or return the portable agent inbox through /api/foundation-agents/identity with action provision_inbox. Use this when an agent needs a reusable email address and inbox identity without forcing dashboard setup.

Register an AI agent on the AgentPay network. Returns an agentId and agentKey that identify the agent. No merchant account required - agents can self-register. The agentKey is shown once; store it securely. Trust score builds automatically after confirmed transactions.

Report a possible provider credential exposure through /api/capabilities/credential-exposure-reports. Never include the credential itself. AgentPay opens a hosted rotate-and-reconnect step so the replacement stays out of agent context.

Request a secure external capability connect session through /api/capabilities/connect-sessions. Use this when a human must attach an API key or external credential without ever handing the raw secret to the agent.

Send a message from the portable agent inbox through /api/foundation-agents/identity with action send_inbox_message. Use this for agent-driven email actions once the inbox is provisioned.

Scan the MCP server's local environment for known API keys that can be vaulted with AgentPay. Returns detected providers (found in process.env) and undetected ones (missing). No keys are transmitted — this is read-only. Call this first, then agentpay_vault_env_keys with the providers the developer wants to vault. After a single 6-digit OTP confirmation, those keys are stored encrypted and usable by agents via agentpay_execute_capability — the developer never manages them again.

Start phone verification through /api/foundation-agents/identity with action start_phone_verification. Use this when an agent needs an SMS or voice challenge started for a phone number.

Vault API keys detected in the local environment into AgentPay's encrypted capability vault. Keys are encrypted in transit and at rest (AES-256-GCM). Agents call them through the proxy and never see raw values. Sends a 6-digit OTP to the developer's registered email. Complete with agentpay_confirm_vault. After confirmation, agents can call agentpay_execute_capability with the returned capabilityIds — no more API key management, no more dashboard logins.

Verify an agent identity through /api/foundation-agents/identity with action verify. Use this when an operator has evidence to attest ownership, environment, and proofs for an agent.

Verify an issued identity credential through /api/foundation-agents/identity with action verify_credential.

Claim a task for execution

Read assigned tasks from Bee board

Query shared vault notes

Report task completion result

Query agent capabilities and state

Get unified fleet state snapshot

Create a payment intent for a merchant (USDC).

Retrieve an AgentPassport for an agent (optional).

Deliver a webhook payload to the AgentPay webhook handler (useful for async notifications).

Confirm TOTP enrollment after scanning the QR. Provide the 6-digit code from your authenticator app.

Enroll TOTP for registry actions that require it (paid subscriptions, publishing). Returns: - setup_url: open in browser to see a QR code to scan - setup_key: manually enter in Google Authenticator / Authy - otpauth_uri: the full otpauth:// URI Free server subscriptions do NOT require TOTP. Only call this when you need to publish or subscribe to a paid server.

List all MCP servers you are subscribed to, with usage counts and plan details.

View your publisher payout history and current month pending earnings. Payouts are 70% of billed revenue, processed monthly.

Publish an MCP server to the AgentPay marketplace. No web UI required. Supported transports: - http: provide endpoint_url (must be https://). Requires domain verification after publishing. Cloudflare Workers (*.workers.dev): use /.well-known/agentpay-publisher.json method. - stdio: provide github_url + command + args. Goes active immediately, no verification needed. Always requires TOTP authentication.

Search the AgentPay MCP server marketplace. Returns available servers with pricing, transport type (http/stdio), and install counts. Free servers: subscribe and use immediately, no TOTP needed. Paid servers: require TOTP (call registry_enroll first). After finding a server: call registry_subscribe, then registry_server_info for harness config.

Get full details and ready-to-paste harness config for a specific MCP server. For HTTP servers: returns URL-based config for Claude Code, Codex, Cursor. For stdio servers: returns command-based config + install instructions. You must be subscribed to see connection config.

Subscribe to an MCP server. Free servers activate immediately — no TOTP needed. Paid servers require TOTP (call registry_enroll first if not enrolled). After subscribing: call registry_server_info for harness connection config.

Usage stats and earnings for MCP servers you have published. Shows installs, calls, and revenue.

Verify domain ownership for an HTTP server after publishing. First place your verification token at: Option A (preferred, works on *.workers.dev): /.well-known/agentpay-publisher.json → {"token":"<your_token>"} Option B: DNS TXT record _agentpay-verify.yourdomain.com = <your_token> The verification_token is returned by registry_publish.

Verify a payment by providing transaction id and hash.