{ "markdown": "# AgentPay\n\n

\n Trust + capability vault + governed paid execution for agents.\n

\n\n

\n \"npm\"\n \"CI\"\n \"Live\n \"License\"\n \"Stars\"\n

\n\n---\n\nAgentPay is the authority layer between an AI agent and the real world.\nIt handles provider access, funding authority, human approval, exact-call resume, and proof so agents can finish work end to end without raw secrets in chat or repeated dashboard setup.\n\nIt is not a generic \"agent platform.\"\nIt is the trust seam that makes autonomous execution usable.\n\n```bash\nnpx -y @agentpayxyz/mcp-server\n```\n\nThat one command gives any MCP-compatible host the ability to:\n\n- request API access\n- vault credentials without exposing the raw key to the agent\n- run governed paid execution\n- pause for OTP or approval only when needed\n- resume the exact blocked call automatically\n- reuse governed access later from the same workbench\n\n---\n\n## AgentPay Commerce Network\n\nAgentPay now has a focused ecommerce surface: shoppers discover products by a concrete need, then set a visual Buyer Constitution for category scope, budget, delivery, returns, merchant policy, and human approval. AgentPay enforces that constitution before optional model ranking, and approved matches continue to merchant checkout. Merchants get the other side of the demand graph through catalog truth, source-grounded product imagery, and signed attribution drafts.\n\n- Shopper experience: `/commerce`\n- Seller growth workspace: `/commerce/studio`\n- Need-led API: `POST /api/commerce/discover`\n- Closed-world GPT-5.6 compiler: `POST /api/commerce/compile`\n- Catalog audit API: `POST /api/commerce/catalog/audit`\n- MCP tools: `agentpay_discover_products`, `agentpay_compile_product_decision`, and `agentpay_audit_catalog_truth`\n- Founder and SEO specification: [docs/COMMERCE_CONTROL_PLANE.md](./docs/COMMERCE_CONTROL_PLANE.md)\n- OpenAI Build Week evidence and judge runbook: [docs/OPENAI_BUILD_WEEK_2026.md](./docs/OPENAI_BUILD_WEEK_2026.md)\n- Dated hackathon reuse decisions and event-specific build boundaries: [docs/HACKATHON_REUSE_2026-07-18.md](./docs/HACKATHON_REUSE_2026-07-18.md)\n\nThe included inventory and demand indexes are sandbox fixtures. They do not represent live merchants, customers, orders, or revenue. Merchant connections, public inventory, conversion webhooks, fee collection, and deployment remain separate approval and validation gates.\n\n---\n\n## Casper Buildathon Entry\n\n**Bee + Clickey: Founder-in-a-Box with Casper-Verifiable Agent Payments**\n\nBee is the operator brain. Clickey is the desktop companion. Together they route work across a real local AI fleet, discover tools through the AgentPay Feed MCP, and stage payment receipts through founder-approved rails. For the Casper Agentic Buildathon, the project adds a Casper-native receipt path: Bee can stage `x402-casper` payment payloads and bind a Bee mandate/receipt id to a Casper Testnet transfer.\n\nStart here:\n\n- [Submission packet](CASPER_AGENTIC_BUILDATHON_SUBMISSION.md)\n- [Post-deadline recovery packet](CASPER_POST_DEADLINE_RECOVERY_PACKET.md)\n- [Bee control plane](ops/mac-mini/bee/README.md)\n- [Casper receipt runner](ops/mac-mini/bee/casper/README.md)\n\nJudges can verify the local proof path:\n\n```bash\nnode --test ops/mac-mini/bee/bee.test.mjs\nops/mac-mini/bin/bee doctor\nops/mac-mini/bin/bee caps\nops/mac-mini/bin/bee feed\nnpm --prefix ops/mac-mini/bee/casper run proof\nnpm --prefix ops/mac-mini/bee/casper run video\n```\n\nBroadcasting a real testnet proof requires a dedicated funded Casper testnet key:\n\n```bash\nCASPER_PRIVATE_KEY_HEX= \\\nCASPER_RECIPIENT_PUBLIC_KEY_HEX= \\\nBEE_CASPER_RECEIPT_ID= \\\nnpm --prefix ops/mac-mini/bee/casper run receipt\n```\n\nBee can prepare and stage high-impact actions, but it does not autonomously publish, upload, change OAuth/account settings, move money, or bypass founder approval. Live settlement remains founder-triggered.\n\n---\n\n## Why it exists\n\nAgents fail at the same place every time:\n\n1. they need an API the developer has not wired yet\n2. they need a credential the human should not paste into chat\n3. they hit a paid step and lose continuity\n\nAgentPay fixes that seam.\n\n**Capability Vault**\nUsers connect a provider once through AgentPay. The raw credential is vaulted server-side. The agent receives only governed capability access.\n\n**Governed paid execution**\nHumans set guardrails once: funding rail, auto-approve limit, OTP policy, and spend limits. AgentPay enforces those rules when the agent acts.\n\n**Exact-call resume**\nWhen a paid step needs a human, AgentPay pauses, collects the minimum approval, and resumes the exact blocked call without asking the agent to reconstruct it.\n\n**Same-workbench reuse**\nLocal projects never need raw provider keys. AgentPay can issue opaque, revocable workbench leases so the same workbench can reuse governed access later.\n\n## The inevitable path\n\nThis is the path AgentPay is built around:\n\n1. Agent asks for an API.\n2. AgentPay checks whether governed access already exists.\n3. If not, AgentPay runs one hosted setup flow for authority and provider connection.\n4. The agent uses the capability for free until paid usage is required.\n5. AgentPay pauses for OTP or confirmation only if policy requires it.\n6. AgentPay resumes the exact blocked call.\n7. The same workbench can reuse governed access later without re-entering the secret.\n\n---\n\n## Get started in 30 seconds\n\nAdd AgentPay to Claude Desktop (`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:\n\n```bash\ncurl -s -X POST https://api.agentpay.so/api/merchants/register \\\n -H \"Content-Type: application/json\" \\\n -d '{ \"name\": \"My Agent\", \"email\": \"you@example.com\" }'\n```\n\nNow ask your host:\n\n> \"My agent needs Firecrawl. Set a $5 auto-approve limit, ask for OTP above that, and keep the key out of chat.\"\n\nOr:\n\n> \"My agent needs Databento for this workbench. If access already exists, reuse it. If not, start the minimal AgentPay setup flow.\"\n\n**[Full quickstart](./QUICKSTART.md)**\n\n---\n\n## Terminal-native control plane\n\nAgentPay should be operated through hosts and terminals, not a merchant dashboard.\n\nKey surfaces:\n\n- `GET /api/capabilities/authority-bootstrap`\n- `POST /api/capabilities/authority-bootstrap`\n- `POST /api/capabilities/access-resolve`\n- `POST /api/capabilities/onboarding-sessions`\n- `POST /api/capabilities/lease-execute`\n- `GET /api/capabilities/leases`\n- `POST /api/capabilities/leases/:leaseId/revoke`\n- `POST /api/capabilities/:capabilityId/execute`\n\nThese let a host or agent:\n\n- read authority state\n- set guardrails\n- connect providers\n- request human approval only when needed\n- reuse governed access safely later\n\n## Flagship provider paths\n\nThe current wedge is strongest when AgentPay owns setup and continuity for high-value agent APIs.\n\nCurrent priority paths:\n\n- Databento\n- Firecrawl\n- Browserbase\n- Exa\n- Generic REST API fallback\n\nThe product goal is simple: visiting provider dashboards should become the exception, not the default.\n\n## Remote MCP\n\nFor hosts that support remote MCP:\n\n```\nhttps://api.agentpay.so/api/mcp\n```\n\nAuthenticate with your API key as a Bearer token, or mint a short-lived token for Claude, OpenAI, or another remote MCP host:\n\n```bash\ncurl -X POST https://api.agentpay.so/api/mcp/tokens \\\n -H \"Authorization: Bearer apk_your_key_here\" \\\n -d '{ \"audience\": \"openai\", \"ttlSeconds\": 3600 }'\n```\n\n## Ace - built on AgentPay\n\n[Ace](apps/meridian/README.md) is a proof front door built on AgentPay.\nIt demonstrates the core seam under real-world conditions. It is not the core story.\n\n## Product truth\n\nIf a human still has to:\n\n- paste a raw provider key into chat\n- rebuild a blocked call after payment\n- keep reopening provider dashboards\n- or lose continuity between approval and execution\n\nthen the product is still unfinished.\n\n## Developer resources\n\n| Resource | Link |\n|----------|------|\n| Quickstart | [QUICKSTART.md](./QUICKSTART.md) |\n| MCP server reference | [packages/mcp-server/README.md](./packages/mcp-server/README.md) |\n| Full API reference | [openapi.yaml](./openapi.yaml) |\n| Terminal-native control plane | [docs/TERMINAL_NATIVE_CONTROL_PLANE_20260419.md](./docs/TERMINAL_NATIVE_CONTROL_PLANE_20260419.md) |\n| Examples | [examples/README.md](./examples/README.md) |\n| Security model | [docs/SECURITY_MODEL.md](./docs/SECURITY_MODEL.md) |\n| Secure payment providers | [docs/PAYMENT_PROVIDERS.md](./docs/PAYMENT_PROVIDERS.md) |\n| Commerce network | [docs/COMMERCE_CONTROL_PLANE.md](./docs/COMMERCE_CONTROL_PLANE.md) |\n\n## Repository layout\n\n```\napps/\n api-edge/ Cloudflare Workers public API\n meridian/ Ace front door\n\npackages/\n mcp-server/ Published MCP package\n sdk/ TypeScript SDK\n sdk-node/ Node.js SDK\n\nexamples/ Example agents and adapters\ndocs/ Architecture and product notes\nmigrations/ PostgreSQL migrations\nops/ Founder and growth operating artifacts\n```\n\n## License\n\nBusiness Source License 1.1. Converts to AGPL-3.0 on 2029-01-01.\n", "bytes": 9982, "sha": "ce6ad1485a41b4cc8eca16110cf33148db631d75bad4de9e616e726d6c273ce1", "repo_slug": "rumblingb/agentpay", "fonte": "repo", "truncated": false, "api": "https://agentalog.com/api/listings/mcp_io_github_rumblingb_agentpay_b0119dd2/readme" }