ArcPay Arbitrum MCP

ArcPay Arbitrum MCP is ArcPay for AI agents. Claude Desktop, Codex-compatible hosts, and other MCP clients can ask ArcPay for Arbitrum deployment data, derive agent and invoice IDs, prepare x402 paid-resource flows, generate privacy/invoice instructions, build GMX/ZeroDev/Dune/Fhenix execution handoff payloads, and return evidence checklists before claiming any work is complete.

Install

npm install -g @arcpaylabs/arbitrum-mcp

Claude Desktop

Add this to your Claude Desktop MCP config:

{
  "mcpServers": {
    "arcpay-arbitrum": {
      "command": "arcpay-arbitrum-mcp"
    }
  }
}

Restart Claude Desktop after editing the config.

Tools

• get_deployment
• derive_agent_id
• derive_invoice_id
• derive_claim_hash
• derive_privacy_commitment
• privacy_intent_guide
• invoice_guide
• x402_guide
• agent_onboarding_payload
• usdc_card_plan
• policy_plan
• evidence_template
• execution_handoff
• gmx_execution_plan
• zerodev_session_policy
• dune_evidence_spec
• fhenix_privacy_boundary
• demo_path
• smoke_commands

Hosted Surfaces

• App: https://arcpay-arbitrum.vercel.app
• Docs: https://arcpay-arbitrum.vercel.app/docs/overview
• OpenAPI: https://arcpay-arbitrum.vercel.app/openapi.json
• llms.txt: https://arcpay-arbitrum.vercel.app/llms.txt
• x402: https://arcpay-arbitrum.vercel.app/api

The MCP server makes ArcPay usable by agents directly, not only by humans clicking a dashboard. It does not sign transactions or mutate treasury state; it returns deterministic IDs, integration guidance, handoff payloads, and public deployment metadata that an operator can verify before execution.

Agent-Native Flows

agent_onboarding_payload lets Claude/Codex/custom agents request the same onboarding payload a dashboard user gets: agent id, x402 endpoint, contract map, policy requirements, optional ERC-8004 identity context, and claim-code steps.

usdc_card_plan lets an agent or developer prepare card issuance without touching the dashboard. It returns the card id, vault/token contracts, call sequence, and proof requirements for create/top-up/spend.

policy_plan returns both global workspace controls and per-agent controls, so an agent can explain what it is allowed to do before attempting any paid, sponsored, GMX, or money-moving action.

evidence_template is the guardrail: it tells the agent exactly what hashes, API responses, GMX/ZeroDev/Dune/Fhenix evidence, and screenshots are required before it can claim completion.