MCP Server

Drop AgentPay into Claude Desktop or any MCP host in 30 seconds. 30+ tools across mandates, credentials, payments, and identity — no infrastructure to build.

Install

npx -y @agentpayxyz/mcp-server

Claude Desktop config

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):

{
  "mcpServers": {
    "agentpay": {
      "command": "npx",
      "args": ["-y", "@agentpayxyz/mcp-server"],
      "env": {
        "AGENTPAY_API_KEY": "apk_your_key_here",
        "AGENTPAY_MERCHANT_ID": "mer_your_merchant_id"
      }
    }
  }
}

Get your API key: POST /api/merchants/register — register with one curl call, no Stripe or wallet needed.

Remote MCP (no local process)

https://api.agentpay.so/api/mcpPass your API key as Authorization: Bearer apk_...

# Mint a short-lived token (useful for OpenAI / Anthropic remote MCP)
curl -X POST https://api.agentpay.so/api/mcp/tokens \
  -H "Authorization: Bearer apk_your_key_here" \
  -H "Content-Type: application/json" \
  -d '{"audience":"openai","ttlSeconds":3600}'

Governed Mandates Capability Vault Payments & Funding Identity & AgentPassport Agent Registry

Governed Mandates

Mandates are the safety layer for autonomous agents. The agent proposes an objective and budget cap; the principal approves once; AgentPay enforces automatically through the full create → plan → approve → execute lifecycle.

agentpay_create_mandate

Create a governed mandate and auto-trigger the planning step. Returns a mandate record with status, AI-generated recommendation, and approval requirement.

param	type	req	description
`principalId`	string	yes	The human principal who owns the mandate and holds approval authority
`operatorId`	string	yes	The agent or operator initiating the mandate
`objective`	string	yes	Outcome-oriented plain-language description of what should happen
`source`	string	no	"direct_human" or "delegated_agent" (default: delegated_agent)
`constraints`	object	no	Execution constraints — e.g. { budgetMax: 5000 } in pence/cents
`mandate`	object	no	Policy — e.g. { amountPence: 5000, currency: "GBP", autoApproveAmountPence: 2000, approvalMethod: "inline" }

agentpay_approve_mandate

Approve a mandate in awaiting_approval state. Transitions to approved so execution can begin.

param	type	req	description
`intentId`	string	yes	Mandate ID to approve
`actorId`	string	no	Operator or principal ID recording who approved
`approvalToken`	string	no	Ephemeral token from a linked approval session

agentpay_execute_mandate

Start execution of an approved mandate. Returns execution state and journey session ID.

param	type	req	description
`intentId`	string	yes	Mandate ID to execute
`actorId`	string	no	Agent/operator identifier starting execution
`jobId`	string	no	Optional downstream execution job ID

agentpay_get_mandate

Fetch the full mandate record including current status, approval state, recommendation, and latest journey state.

param	type	req	description
`intentId`	string	yes	Mandate ID to fetch

agentpay_get_mandate_journey_status

Fetch the live journey or execution session status for a mandate that is currently executing. If no journey session exists yet, that is expected until downstream dispatch starts.

param	type	req	description
`intentId`	string	yes	Mandate ID

agentpay_get_mandate_history

Fetch the append-only audit timeline for a mandate — all state transitions, approvals, and execution events.

param	type	req	description
`intentId`	string	yes	Mandate ID

agentpay_cancel_mandate

Cancel or revoke a mandate that has not yet started executing. Audit trail is preserved.

param	type	req	description
`intentId`	string	yes	Mandate ID to cancel
`actorId`	string	no	Who is cancelling — for audit trail
`reason`	string	no	Optional cancellation reason

Capability Vault

The Capability Vault lets agents call external APIs (Firecrawl, Perplexity, OpenAI, Browserbase…) without ever seeing the raw API key. The user vaults the key once via a secure connect flow; all future calls proxy through AgentPay.

agentpay_list_capability_providers

List the governed external capability catalog — all providers AgentPay can vault and proxy.

agentpay_request_capability_connect

Start a secure connect session for a capability provider. Returns a connectUrl the user opens to vault their API key. The agent never sees the raw credential.

param	type	req	description
`provider`	string	yes	Provider slug — firecrawl, perplexity, openai, tavily, exa, browserbase, serper, google_maps, etc.
`capabilityKey`	string	yes	Stable merchant-scoped identifier for this capability, e.g. "firecrawl_primary"
`subjectType`	string	yes	"merchant", "principal", "agent", or "workspace"
`subjectRef`	string	yes	ID of the owning merchant, principal, agent, or workspace
`baseUrl`	string	no	Base URL — implied by provider preset if omitted
`allowedHosts`	string[]	no	Allowed upstream hosts — implied by preset if omitted
`authScheme`	string	no	"bearer", "x_api_key", or "basic" — implied by preset
`credentialKind`	string	no	"api_key", "bearer_token", or "basic_auth" — implied by preset
`freeCalls`	number	no	Free-call quota before paid usage approval is required

agentpay_list_capabilities

List all capabilities already connected for your merchant account.

param	type	req	description
`merchantId`	string	yes	Your merchant ID

agentpay_get_capability

Fetch metadata for a single connected capability.

param	type	req	description
`capabilityId`	string	yes	Capability ID to fetch

agentpay_execute_capability

Execute an upstream API call through the governed proxy. AgentPay injects the vaulted credential, enforces the allowlisted host, and returns approval_required once the spend allowance is exhausted.

param	type	req	description
`capabilityId`	string	yes	Connected capability ID
`endpoint`	string	yes	Upstream path to call (e.g. /v1/scrape)
`method`	string	no	HTTP method (default: POST)
`body`	object	no	Request body forwarded to the upstream API

Payments & Funding

Tools for inline human funding — Stripe Checkout (card) and Razorpay (UPI) — and for vaulting reusable payment methods so governed mandates can charge without re-prompting.

agentpay_create_human_funding_request

Create a host-native human funding request. Returns a nextAction payload with a Stripe Checkout URL (card) or UPI QR/deep-link the host can render inline. The human pays without leaving the chat.

param	type	req	description
`amount`	number	yes	Amount to request
`currency`	string	yes	Currency code (GBP, USD, INR…)
`description`	string	no	What this payment is for
`merchantId`	string	yes	Your merchant ID
`preferredMethod`	string	no	card or upi (default: card)

agentpay_create_funding_setup_intent

Create a reusable Stripe funding setup intent so the principal can save a card for future governed spending.

param	type	req	description
`principalId`	string	yes	Principal/user ID to attach the card to
`merchantId`	string	yes	Your merchant ID

agentpay_confirm_funding_setup

Confirm a completed Stripe funding setup and persist the saved payment method.

param	type	req	description
`setupIntentId`	string	yes	Stripe Setup Intent ID to confirm

agentpay_list_funding_methods

List all saved funding methods for a principal.

param	type	req	description
`principalId`	string	yes	Principal ID

agentpay_create_payment_intent

Create a USDC payment intent and return a Solana Pay URI.

param	type	req	description
`amount`	number	yes	Amount in USD
`description`	string	no	Payment description
`merchantId`	string	yes	Your merchant ID

agentpay_get_intent_status

Check whether a payment has been confirmed on-chain.

param	type	req	description
`intentId`	string	yes	Payment intent ID

agentpay_get_receipt

Get the full settlement receipt for a completed payment.

param	type	req	description
`intentId`	string	yes	Payment or mandate ID

agentpay_parse_upi_payment_request

Parse and normalize a raw upi://pay URI or decoded QR payload. Returns payee, amount, reference, and all standard fields.

param	type	req	description
`upiUri`	string	yes	Raw upi://pay URI or decoded QR string

Identity & AgentPassport

AgentPassport is portable agent identity — trust score, linked accounts, phone verification, and credential attestations that travel across platforms.

agentpay_get_passport

Look up an agent's trust score, interaction history, and linked identities.

param	type	req	description
`agentId`	string	yes	Agent ID (agt_… or string identifier)

agentpay_get_identity_bundle

Fetch the portable identity bundle for an agent — verified identities, linked accounts, and trust state.

param	type	req	description
`agentId`	string	yes	Agent ID

agentpay_verify_identity_bundle

Verify an agent identity and issue a credential.

param	type	req	description
`agentId`	string	yes	Agent ID to verify

agentpay_verify_identity_credential

Verify an already-issued identity credential.

param	type	req	description
`credentialId`	string	yes	Credential ID to verify

agentpay_link_identity_bundles

Link two agent identities together in the trust graph.

param	type	req	description
`agentId`	string	yes	Primary agent ID
`linkedAgentId`	string	yes	Agent ID to link

agentpay_start_identity_phone_verification

Start a phone verification challenge for an agent identity.

param	type	req	description
`agentId`	string	yes	Agent ID
`phoneNumber`	string	yes	E.164 phone number

agentpay_confirm_identity_phone_verification

Confirm a phone verification challenge with the OTP code.

param	type	req	description
`agentId`	string	yes	Agent ID
`verificationId`	string	yes	Verification ID from start step
`code`	string	yes	OTP code from SMS

agentpay_provision_identity_inbox

Provision or return a portable agent inbox. Returns an inbox address the agent can use to send and receive messages across platforms.

param	type	req	description
`agentId`	string	yes	Agent ID

agentpay_send_identity_inbox_message

Send a message from a provisioned agent inbox.

param	type	req	description
`agentId`	string	yes	Sending agent ID
`to`	string	yes	Recipient address or agent ID
`body`	string	yes	Message body

agentpay_list_identity_inbox_messages

List recent inbox messages for an agent.

param	type	req	description
`agentId`	string	yes	Agent ID
`limit`	number	no	Number of messages to return (default: 10)

Agent Registry

agentpay_register_agent

param	type	req	description
`name`	string	yes	Agent display name
`description`	string	no	What this agent does
`merchantId`	string	yes	Owning merchant ID

agentpay_get_agent

Look up a registered agent's public identity record.

param	type	req	description
`agentId`	string	yes	Agent ID to look up

agentpay_discover_agents

Search and filter agents on the network by capability, trust level, or name.

param	type	req	description
`query`	string	no	Search query
`capability`	string	no	Filter by capability (e.g. travel, finance)
`limit`	number	no	Results to return (default: 20)

agentpay_get_merchant_stats

Get your account's payment and mandate statistics.

param	type	req	description
`merchantId`	string	yes	Your merchant ID

Environment variables

variable	required	description
`AGENTPAY_API_KEY`	yes	Your merchant API key (apk_…)
`AGENTPAY_MERCHANT_ID`	no	Your merchant ID (mer_…). Recommended — used in mandate and capability calls.
`AGENTPAY_API_URL`	no	Override the API base URL. Defaults to the production Workers endpoint.

Example prompts

Once configured, ask Claude any of these to exercise the full tool surface:

> “Create a governed mandate to book a train London to Bristol, budget £100, require my approval above £50.”
> “Connect Firecrawl without exposing the raw API key to the agent.”
> “Run a Firecrawl scrape through the connected capability.”
> “Create a £49 card payment request so the human can pay without leaving the chat.”
> “Create a ₹499 UPI funding request.”
> “Look up the AgentPassport for agent_001.”
> “Provision an inbox for agent_001 and send a message to traveler@example.com.”
> “List all mandates and their current status.”
> “Show me the audit timeline for mandate mnd_abc123.”

Quickstart →

MCP in 2 min, REST in 5 min

Adapters →

OpenAI, LangChain, Vercel AI SDK

AgentPassport →

Portable identity and trust

Pricing →

Launch free · Builder $39 · Growth $149