{"_id":"@agentpayxyz/mcp-server","_rev":"4-2f2ab38a9da7b2447cc0f77686c0671d","name":"@agentpayxyz/mcp-server","dist-tags":{"latest":"0.2.1"},"versions":{"0.1.0":{"name":"@agentpayxyz/mcp-server","version":"0.1.0","keywords":["agentpay","mcp","model-context-protocol","ai-agents","payments","claude","anthropic","agentpassport"],"author":{"name":"AgentPay"},"license":"BSL-1.1; AGPL-3.0 after 2029-01-01","_id":"@agentpayxyz/mcp-server@0.1.0","maintainers":[{"name":"rumblingb","email":"vishar.rumbling@gmail.com"}],"homepage":"https://github.com/Rumblingb/Agentpay/tree/main/packages/mcp-server#readme","bugs":{"url":"https://github.com/Rumblingb/Agentpay/issues"},"bin":{"agentpay-mcp":"dist/index.js"},"dist":{"shasum":"f69a7025c6e18ba11eb6f07a70c2f2d7382cd993","tarball":"https://registry.npmjs.org/@agentpayxyz/mcp-server/-/mcp-server-0.1.0.tgz","fileCount":6,"integrity":"sha512-18ieLtplKBwFGh5o1oDWKC+jniGIBIJou1axqivTyJO/pjSnEzpIOttxZsl76DL8DE3CULFeia26OykcLRpx/Q==","signatures":[{"sig":"MEYCIQDPkO+REFLbf+gur2cBc0uxkL/MceoQ2iYK6N+KOZGsUAIhAKmofLgbamuXdR3gFpRY1Xv/JfKQ4FP5CiiY4+VhPgXq","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":20912},"main":"./dist/index.js","type":"module","types":"./dist/index.d.ts","engines":{"node":">=20.0.0"},"exports":{".":{"types":"./dist/index.d.ts","import":"./dist/index.js"}},"gitHead":"a4e5da74fe175d8dc8fa7a85edc1464322c555e5","scripts":{"dev":"node --watch dist/index.js","build":"tsc","typecheck":"tsc --noEmit"},"_npmUser":{"name":"rumblingb","email":"vishar.rumbling@gmail.com"},"repository":{"url":"git+https://github.com/Rumblingb/Agentpay.git","type":"git","directory":"packages/mcp-server"},"_npmVersion":"11.9.0","description":"Model Context Protocol (MCP) server for AgentPay — gives AI assistants the ability to create payment intents, verify settlements, and look up AgentPassports","directories":{},"_nodeVersion":"24.14.0","dependencies":{"@modelcontextprotocol/sdk":"^1.0.0"},"_hasShrinkwrap":false,"devDependencies":{"typescript":"^5.3.3","@types/node":"^20.0.0"},"_npmOperationalInternal":{"tmp":"tmp/mcp-server_0.1.0_1773790675115_0.0396668295907876","host":"s3://npm-registry-packages-npm-production"}},"0.1.1":{"name":"@agentpayxyz/mcp-server","version":"0.1.1","keywords":["agentpay","mcp","model-context-protocol","ai-agents","payments","claude","anthropic","agentpassport"],"author":{"name":"AgentPay"},"license":"BSL-1.1; AGPL-3.0 after 2029-01-01","_id":"@agentpayxyz/mcp-server@0.1.1","maintainers":[{"name":"rumblingb","email":"vishar.rumbling@gmail.com"}],"homepage":"https://github.com/Rumblingb/Agentpay/tree/main/packages/mcp-server#readme","bugs":{"url":"https://github.com/Rumblingb/Agentpay/issues"},"bin":{"agentpay-mcp":"bin/agentpay-mcp.mjs"},"dist":{"shasum":"50ea779c8a938c2242b335e0d7e51d29720de199","tarball":"https://registry.npmjs.org/@agentpayxyz/mcp-server/-/mcp-server-0.1.1.tgz","fileCount":7,"integrity":"sha512-BT+U00uZcj/WMSEhy2WmMc9U07uLU+7kPUpwqkEQw9Y5h3g5WMxX9k/5NSwXZUWeawOeBtFVl8UpXNH0zX6QaA==","signatures":[{"sig":"MEYCIQCVt4c8s6SaBFpjk+yAsEZsfBxP0AK7JckziyfH1REMtAIhAMnAE4fnUdUjIlJ9ZuPoBiZwiGJM2WaU61GGWP1/RbNi","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":20973},"main":"./dist/index.js","type":"module","types":"./dist/index.d.ts","engines":{"node":">=20.0.0"},"exports":{".":{"types":"./dist/index.d.ts","import":"./dist/index.js"}},"gitHead":"a4e5da74fe175d8dc8fa7a85edc1464322c555e5","scripts":{"dev":"node --watch dist/index.js","build":"tsc","typecheck":"tsc --noEmit"},"_npmUser":{"name":"rumblingb","email":"vishar.rumbling@gmail.com"},"repository":{"url":"git+https://github.com/Rumblingb/Agentpay.git","type":"git","directory":"packages/mcp-server"},"_npmVersion":"11.9.0","description":"Model Context Protocol (MCP) server for AgentPay — gives AI assistants the ability to create payment intents, verify settlements, and look up AgentPassports","directories":{},"_nodeVersion":"24.14.0","dependencies":{"@modelcontextprotocol/sdk":"^1.0.0"},"_hasShrinkwrap":false,"devDependencies":{"typescript":"^5.3.3","@types/node":"^20.0.0"},"_npmOperationalInternal":{"tmp":"tmp/mcp-server_0.1.1_1773790816967_0.28881716894101306","host":"s3://npm-registry-packages-npm-production"}},"0.2.0":{"name":"@agentpayxyz/mcp-server","version":"0.2.0","keywords":["agentpay","mcp","model-context-protocol","ai-agents","payments","claude","anthropic","agentpassport"],"author":{"name":"AgentPay"},"license":"BSL-1.1; AGPL-3.0 after 2029-01-01","_id":"@agentpayxyz/mcp-server@0.2.0","maintainers":[{"name":"rumblingb","email":"vishar.rumbling@gmail.com"}],"homepage":"https://github.com/Rumblingb/Agentpay/tree/main/packages/mcp-server#readme","bugs":{"url":"https://github.com/Rumblingb/Agentpay/issues"},"bin":{"agentpay-mcp":"bin/agentpay-mcp.mjs"},"dist":{"shasum":"500a5b5852cb71f1edc555663a266f0f36f7e366","tarball":"https://registry.npmjs.org/@agentpayxyz/mcp-server/-/mcp-server-0.2.0.tgz","fileCount":12,"integrity":"sha512-c+y6eWTDS04n6Ilacdmngtyv2JywZNLrn5df4MpjKDpds8pOT63MgdPsQ/17ej87bLJryvxge7UIRmChjPhWcA==","signatures":[{"sig":"MEQCIHEE7eeKK0lbpT9KwkMBMWDZebZp37GX6AyqHZ83PDX8AiBUFkxP7MKuqMeQ71DMGeelq4kBsO65QEI9PD9cYXmhxQ==","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"}],"unpackedSize":161241},"main":"./dist/index.js","type":"module","types":"./dist/index.d.ts","engines":{"node":">=20.0.0"},"exports":{".":{"types":"./dist/index.d.ts","import":"./dist/index.js"}},"mcpName":"io.github.Rumblingb/agentpay","scripts":{"dev":"node --watch dist/index.js","build":"tsc","typecheck":"tsc --noEmit"},"_npmUser":{"name":"rumblingb","email":"vishar.rumbling@gmail.com"},"repository":{"url":"git+https://github.com/Rumblingb/Agentpay.git","type":"git","directory":"packages/mcp-server"},"_npmVersion":"11.9.0","description":"Model Context Protocol (MCP) server for AgentPay — gives AI assistants the ability to create payment intents, verify settlements, and look up AgentPassports","directories":{},"_nodeVersion":"24.14.0","dependencies":{"@modelcontextprotocol/sdk":"^1.0.0"},"_hasShrinkwrap":false,"devDependencies":{"typescript":"^5.3.3","@types/node":"^20.0.0"},"_npmOperationalInternal":{"tmp":"tmp/mcp-server_0.2.0_1776564022084_0.6412718252451715","host":"s3://npm-registry-packages-npm-production"}},"0.2.1":{"_id":"@agentpayxyz/mcp-server@0.2.1","bin":{"agentpay-mcp":"bin/agentpay-mcp.mjs"},"bugs":{"url":"https://github.com/Rumblingb/Agentpay/issues"},"dist":{"shasum":"ed77a941021f77dc068c16998d1f9eb7dadd76c3","tarball":"https://registry.npmjs.org/@agentpayxyz/mcp-server/-/mcp-server-0.2.1.tgz","fileCount":20,"integrity":"sha512-8WaJ9J2KMStu4xuTNq86tiVaUBYZiqUIpPX3HpWo9HSP1KOGuW/pB6pRJemG7er1AYXeAVqPlXZyb+TfXjR2HQ==","signatures":[{"sig":"MEYCIQCWVySwHVSW1RXVT5w8OJppBBOBgYcuhSK+IJuiYqvzEAIhAJs3UaLqYARYa7rn3nLpD3hmvC536G8D37O9Dv403o4M","keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U"},{"keyid":"SHA256:DhQ8wR5APBvFHLF/+Tc+AYvPOdTpcIDqOhxsBHRwC7U","sig":"MEQCIBcBTXeGLmvV4kqy4OciUqFMBtjb2p7reDejBwgj1qUIAiBEyNoYIYbNZiu/1uvYLfPhosfYoHtyW0quLwwWje4tSA=="}],"unpackedSize":209679},"main":"./dist/index.js","name":"@agentpayxyz/mcp-server","type":"module","types":"./dist/index.d.ts","author":{"name":"AgentPay"},"engines":{"node":">=20.0.0"},"exports":{".":{"types":"./dist/index.d.ts","import":"./dist/index.js"}},"gitHead":"c37b0c9fc43015e1aac1f2adab2d5c9bb2eb46de","license":"BSL-1.1; AGPL-3.0 after 2029-01-01","mcpName":"io.github.Rumblingb/agentpay","scripts":{"dev":"node --watch dist/index.js","build":"tsc","typecheck":"tsc --noEmit"},"version":"0.2.1","_npmUser":{"name":"rumblingb","email":"vishar.rumbling@gmail.com"},"homepage":"https://github.com/Rumblingb/Agentpay/tree/main/packages/mcp-server#readme","keywords":["agentpay","mcp","model-context-protocol","ai-agents","payments","claude","anthropic","agentpassport","api-key-management","api-key-vault","credential-management","capability-vault","mandate","agent-billing","agent-payments","spending-limits","zero-api-key","autonomous-agents","agent-infrastructure","mcp-payments","stripe-agent","upi","firecrawl","perplexity","openai-proxy","agent-identity","agentpassport","rotate-api-key","secure-credentials","governed-agent"],"repository":{"url":"git+https://github.com/Rumblingb/Agentpay.git","type":"git","directory":"packages/mcp-server"},"_npmVersion":"10.9.7","description":"MCP server for AgentPay — autonomous agent infrastructure for payments, API key vaulting, and governed mandates. One OTP vaults credentials so agents never handle raw API keys. Agents spend money within user-defined limits. Works with Claude, GPT-4o, and ","directories":{},"maintainers":[{"name":"rumblingb","email":"vishar.rumbling@gmail.com"}],"_nodeVersion":"22.14.0","dependencies":{"@modelcontextprotocol/sdk":"^1.27.1"},"_hasShrinkwrap":false,"devDependencies":{"typescript":"^5.3.3","@types/node":"^20.0.0"},"_npmOperationalInternal":{"host":"s3://npm-registry-packages-npm-production","tmp":"tmp/mcp-server_0.2.1_1788901204239_0.1507424186417219"}}},"time":{"created":"2026-03-17T23:37:55.052Z","modified":"2026-09-08T21:00:04.538Z","0.1.0":"2026-03-17T23:37:55.286Z","0.1.1":"2026-03-17T23:40:17.101Z","0.2.0":"2026-04-19T02:00:22.239Z","0.2.1":"2026-09-08T21:00:04.317Z"},"bugs":{"url":"https://github.com/Rumblingb/Agentpay/issues"},"author":{"name":"AgentPay"},"license":"BSL-1.1; AGPL-3.0 after 2029-01-01","homepage":"https://github.com/Rumblingb/Agentpay/tree/main/packages/mcp-server#readme","keywords":["agentpay","mcp","model-context-protocol","ai-agents","payments","claude","anthropic","agentpassport","api-key-management","api-key-vault","credential-management","capability-vault","mandate","agent-billing","agent-payments","spending-limits","zero-api-key","autonomous-agents","agent-infrastructure","mcp-payments","stripe-agent","upi","firecrawl","perplexity","openai-proxy","agent-identity","agentpassport","rotate-api-key","secure-credentials","governed-agent"],"repository":{"url":"git+https://github.com/Rumblingb/Agentpay.git","type":"git","directory":"packages/mcp-server"},"description":"MCP server for AgentPay — autonomous agent infrastructure for payments, API key vaulting, and governed mandates. One OTP vaults credentials so agents never handle raw API keys. Agents spend money within user-defined limits. Works with Claude, GPT-4o, and ","maintainers":[{"name":"rumblingb","email":"vishar.rumbling@gmail.com"}],"readme":"# @agentpayxyz/mcp-server\n\nModel Context Protocol server for AgentPay. Gives any MCP-compatible AI assistant the ability to create payment intents, verify settlements, look up AgentPassports, manage the portable identity bundle, discover agents, and drive governed mandates on the canonical `/api/mandates` surface.\n\n## Install\n\n```bash\nnpx -y @agentpayxyz/mcp-server\n```\n\n## Hosted Remote MCP\n\nIf you are connecting AgentPay through a host that supports remote MCP, use the deployed endpoint instead of stdio:\n\n- MCP endpoint: `https://api.agentpay.so/api/mcp`\n- Discovery info: `https://api.agentpay.so/api/mcp/info`\n- Pricing metadata: `https://api.agentpay.so/api/mcp/pricing`\n- Host setup guide: `https://api.agentpay.so/api/mcp/setup`\n- Demo flow: `https://api.agentpay.so/api/mcp/demo`\n- Token mint endpoint: `https://api.agentpay.so/api/mcp/tokens`\n\nYou can authenticate with either:\n\n```text\nAuthorization: Bearer apk_your_key_here\n```\n\nor a short-lived hosted MCP token minted from your AgentPay API key:\n\n```bash\ncurl -X POST https://api.agentpay.so/api/mcp/tokens \\\n  -H \"Authorization: Bearer apk_your_key_here\" \\\n  -H \"Content-Type: application/json\" \\\n  -d '{\"audience\":\"openai\",\"ttlSeconds\":3600}'\n```\n\nThe response includes an `accessToken` you can pass to OpenAI or Anthropic as the remote MCP bearer token.\n\nHosted MCP is priced as a headless control plane:\n- `Launch`: free developer tier for evaluation\n- `Builder`: $39/month, includes hosted MCP usage\n- `Growth`: $149/month, higher included volume and lower overage\n- `Funded actions`: 0.75% when AgentPay brokers the human funding step, plus underlying processor fees\n\nSee the public pricing metadata from `GET /api/mcp/pricing`, the host setup guide at `GET /api/mcp/setup`, and the north-star direction in `docs/AGENT_NATIVE_NORTH_STAR.md`.\nMerchants can inspect current hosted MCP charges at `GET /api/mcp/billing/current` and create a payable checkout session at `POST /api/mcp/billing/checkout`.\n\n## Claude Desktop Setup\n\nAdd to your `claude_desktop_config.json`:\n\n```json\n{\n  \"mcpServers\": {\n    \"agentpay\": {\n      \"command\": \"npx\",\n      \"args\": [\"-y\", \"@agentpayxyz/mcp-server\"],\n      \"env\": {\n        \"AGENTPAY_API_KEY\": \"apk_your_key_here\",\n        \"AGENTPAY_MERCHANT_ID\": \"your_merchant_id\"\n      }\n    }\n  }\n}\n```\n\nGet your API key: register at `https://api.agentpay.so/api/merchants/register`\n\n## Tools Exposed\n\n| Tool | Description |\n|------|-------------|\n| `agentpay_create_payment_intent` | Create a USDC payment intent and return a Solana Pay URI |\n| `agentpay_get_intent_status` | Check whether a payment has been confirmed on-chain |\n| `agentpay_get_receipt` | Get the full settlement receipt for a payment |\n| `agentpay_evaluate_purchase` | Apply a Buyer Constitution to researched candidates and return a signed portable Choice Receipt before any mandate or payment |\n| `agentpay_discover_products` | Enforce an explicit Buyer Constitution, then rank caller-supplied catalog candidates with transparent fit factors, sponsorship disclosure, provenance warnings, and a signed draft attribution record |\n| `agentpay_compile_product_decision` | Enforce category, merchant, budget, refund, delivery, and approval rules before GPT-5.6; verify its closed-world ranking and return a signed packet or deterministic fallback without exposing product identity, checkout, or sponsorship data to the model |\n| `agentpay_audit_catalog_truth` | Detect product-data drift across search, agent, protocol, and checkout channel snapshots |\n| `agentpay_parse_upi_payment_request` | Parse and normalize a raw `upi://pay` URI or decoded QR payload through `/api/payments/upi/parse` |\n| `agentpay_get_passport` | Look up an agent's trust score and interaction history |\n| `agentpay_get_identity_bundle` | Fetch the portable identity bundle through `/api/foundation-agents/identity` |\n| `agentpay_verify_identity_bundle` | Verify an agent identity through `/api/foundation-agents/identity` |\n| `agentpay_provision_identity_inbox` | Provision or return a portable agent inbox through `/api/foundation-agents/identity` |\n| `agentpay_send_identity_inbox_message` | Send from a provisioned agent inbox through `/api/foundation-agents/identity` |\n| `agentpay_list_identity_inbox_messages` | List recent inbox messages through `/api/foundation-agents/identity` |\n| `agentpay_start_identity_phone_verification` | Start a phone verification challenge through `/api/foundation-agents/identity` |\n| `agentpay_confirm_identity_phone_verification` | Confirm a phone verification challenge through `/api/foundation-agents/identity` |\n| `agentpay_link_identity_bundles` | Link multiple agent identities through `/api/foundation-agents/identity` |\n| `agentpay_verify_identity_credential` | Verify an issued identity credential through `/api/foundation-agents/identity` |\n| `agentpay_create_funding_setup_intent` | Create a reusable Stripe funding setup intent through `/api/payments/setup-intent` |\n| `agentpay_confirm_funding_setup` | Confirm a completed Stripe funding setup through `/api/payments/confirm-setup` |\n| `agentpay_list_funding_methods` | List saved principal funding methods through `/api/payments/methods/:principalId` |\n| `agentpay_create_human_funding_request` | Create a host-native human funding request through `/api/payments/funding-request` with card-first funding and UPI fallback |\n| `agentpay_list_capability_providers` | List the governed external capability catalog through `/api/capabilities/providers/catalog` |\n| `agentpay_request_capability_connect` | Start a secure external capability connect session through `/api/capabilities/connect-sessions` |\n| `agentpay_list_capabilities` | List connected external capabilities through `/api/capabilities` |\n| `agentpay_get_capability` | Fetch one connected external capability through `/api/capabilities/:capabilityId` |\n| `agentpay_execute_capability` | Execute an upstream API call through the governed proxy at `/api/capabilities/:capabilityId/execute` |\n| `agentpay_create_mandate` | Create and plan a governed mandate on `/api/mandates` |\n| `agentpay_get_mandate` | Fetch the full mandate record, including approval and execution state |\n| `agentpay_get_mandate_journey_status` | Fetch the current journey/execution session for a mandate; `No journey session found` is expected until downstream dispatch starts |\n| `agentpay_get_mandate_history` | Fetch the audit timeline for a mandate from `/api/mandates/:intentId/history` |\n| `agentpay_approve_mandate` | Approve a mandate through `/api/mandates/:intentId/approve` |\n| `agentpay_execute_mandate` | Start execution through `/api/mandates/:intentId/execute` |\n| `agentpay_cancel_mandate` | Cancel or revoke a non-executing mandate through `/api/mandates/:intentId/cancel` |\n| `agentpay_discover_agents` | Search and filter agents on the network |\n| `agentpay_get_merchant_stats` | Get your account's payment statistics |\n| `agentpay_register_agent` | Register an AI agent on the network |\n| `agentpay_get_agent` | Look up a registered agent's public identity record |\n\n## Example Claude Prompt\n\nOnce configured, you can ask Claude:\n\n> \"Given these catalog candidates, rank bags and clothing for a rain-ready commute under GBP 150 that arrive within three days and allow at least 30-day returns. Put those rules in my Buyer Constitution, require my approval, disclose sponsorship, and do not relax any limit.\"\n\nClaude will call `agentpay_discover_products`, return only eligible supplied candidates, show the fit factors and provenance warning, and preserve the supplied checkout plus draft attribution evidence. It will not treat the report as proof of live inventory, a merchant connection, or a fee agreement.\n\n> \"Compile the safest shortlist under that Buyer Constitution. Do not add a category, merchant, or product that did not pass it.\"\n\nClaude will call `agentpay_compile_product_decision`. AgentPay runs the hard filters first, replaces product IDs with opaque references, sends GPT-5.6 only structured numeric evidence, verifies the returned references and rationale codes, and falls back to deterministic rank if the model invents or omits anything. The tool does not create a mandate or payment.\n\n> \"Audit this product across its landing page, Google Merchant, OpenAI feed, UCP, and checkout before recommending it.\"\n\nClaude will call `agentpay_audit_catalog_truth` and return exact drift, freshness, shipping, returns, and identifier issues.\n\n> \"Create a governed mandate to book a train from London to Manchester, and show me the current mandate state.\"\n\nClaude will call `agentpay_create_mandate`, then `agentpay_get_mandate`, and return the mandate record directly.\n\n> \"Create a $5 USDC payment request for a research task and give me the payment link.\"\n\nClaude will call `agentpay_create_payment_intent` and return the Solana Pay URI directly.\n\n> \"Parse this UPI payment request and tell me who it pays, how much it requests, and any reference fields: `upi://pay?pa=merchant@upi&pn=Demo%20Store&am=499.00&cu=INR&tn=Expo%20booth`\"\n\nClaude will call `agentpay_parse_upi_payment_request` and return the normalized UPI request fields directly.\n\n> \"Look up the AgentPassport for `agent_001`.\"\n\nClaude will call `agentpay_get_passport` and summarize the trust score and history.\n\n> \"Get the portable identity bundle for `agent_001`.\"\n\nClaude will call `agentpay_get_identity_bundle` against `/api/foundation-agents/identity` and return the verified identity record, linked identities, and trust state.\n\n> \"Verify this agent identity and return the credential.\"\n\nClaude will call `agentpay_verify_identity_bundle` and return the issued credential and trust level.\n\n> \"Provision an inbox for `agent_001` and show me the address.\"\n\nClaude will call `agentpay_provision_identity_inbox` and return the portable inbox identity.\n\n> \"Send `traveler@example.com` a handled-trip email from `agent_001`.\"\n\nClaude will call `agentpay_send_identity_inbox_message` and return the message result.\n\n> \"Show me the last 10 inbox messages for `agent_001`.\"\n\nClaude will call `agentpay_list_identity_inbox_messages` and return the recent inbox timeline.\n\n> \"Start phone verification for `agent_001` at `+447700900123`.\"\n\nClaude will call `agentpay_start_identity_phone_verification` and return the verification challenge details.\n\n> \"Confirm phone verification for `agent_001` with verification ID `verif_123` and code `123456`.\"\n\nClaude will call `agentpay_confirm_identity_phone_verification` and return the confirmed phone verification state.\n\n> \"Create a reusable funding setup for `principal_001` so my agent can save a card for governed spending.\"\n\nClaude will call `agentpay_create_funding_setup_intent` and return the Stripe setup intent details.\n\n> \"List the funding methods saved for `principal_001`.\"\n\nClaude will call `agentpay_list_funding_methods` and return the saved payment methods.\n\n> \"Create a UPI funding request for ₹499 so the human can pay without leaving the chat.\"\n\nClaude will call `agentpay_create_human_funding_request` and return a `nextAction` payload with a short URL and UPI QR/deep-link data the host can render inline.\n\n> \"Create a card funding request for GBP 49 so the human can pay without leaving the chat.\"\n\nClaude will call `agentpay_create_human_funding_request` and return a `nextAction` payload with a Stripe Checkout URL the host can render inline or open as a scoped secure step.\n\n> \"Connect Firecrawl without exposing the raw API key to the agent.\"\n\nClaude will call `agentpay_request_capability_connect` and return an `auth_required` connect session. The human completes the secure step, AgentPay vaults the key, and future calls use only the returned capability reference.\n\n> \"Run this scrape through the connected Firecrawl capability and ask before spending past the free tier.\"\n\nClaude will call `agentpay_execute_capability`. AgentPay injects the vaulted credential, enforces the allow-listed host, and returns `approval_required` once the free-call allowance is exhausted.\n\n## Environment Variables\n\n| Variable | Required | Description |\n|----------|----------|-------------|\n| `AGENTPAY_API_KEY` | Yes | Your merchant API key |\n| `AGENTPAY_MERCHANT_ID` | Recommended | Your merchant ID (used in intent creation) |\n| `AGENTPAY_API_URL` | No | Override API URL (default: `https://api.agentpay.so`) |\n\n## Publish `@agentpayxyz/mcp-server` 0.2.1\n\nThis repo does **not** auto-publish the MCP package. Do not publish from a cloud agent unless you have an npm token and an explicit release request.\n\nAfter this PR is merged:\n\n```bash\ncd packages/mcp-server\nnpm run build\nnpm pack --dry-run   # confirm bin/agentpay-mcp.mjs and dist/ are included\nnpm publish --access public\n```\n\nConfirm the published tarball boots:\n\n```bash\nnpx -y @agentpayxyz/mcp-server@0.2.1 --help\n```\n\nA missing `AGENTPAY_API_KEY` may warn on stderr. The process must stay up and speak MCP stdio instead of exiting 0.\n","readmeFilename":"README.md"}