Developers increasingly run coding agents that need live infrastructure—spin up a US residential endpoint, verify quota, paste credentials into an integration test. JoyProxy exposes two AI-facing surfaces: OpenClaw Skill (natural language) and AI MCP (structured tools for IDEs).
What both integrations share
- Authentication via an AI Access Token from the dashboard AI Extraction Console.
- Read-only operations — extract proxies, read balances/usage; no arbitrary account mutations.
- Same upstream API as manual extraction—agents automate what you could click, not a hidden tier.
OpenClaw Skill
Skill endpoint: https://api.joyproxy.com/Skill. After connecting your token, you can prompt for region-based short-term extraction or ask usage questions in plain English. Best for operators who already use OpenClaw for task automation.
Product page: OpenClaw Skill.
AI MCP
MCP endpoint: https://api.joyproxy.com/Mcp. Paste the JSON config into Cursor, VS Code, or Claude Desktop so tools appear in the agent sidebar. Best when you want deterministic tool calls in a repo workflow.
Product page: AI MCP.
Safety practices
- Issue a dedicated AI token; revoke it when the experiment ends.
- Do not commit tokens to git—use local env or secret stores.
- Validate agent-generated proxy hosts before pointing production scrapers at them.
AI Customer Service (24/7 chat) remains free for onboarding questions if you prefer human escalation.