AgentPhone-AI/agentphone-mcp — reverse-engineered prompt

Reverse engineered prompt

Build me an MCP server for AgentPhone so my AI assistant can use real phone numbers, send and receive texts, and place voice calls from Cursor, Claude Desktop, or any other MCP client. I want the main idea to be “agents”, where each agent has its own phone number, voice, prompt, and optional webhook, so I can set up things like a support agent, sales agent, or scheduling assistant.

It should let me buy and manage numbers, create and update agents, switch voices, send messages, view conversations, place calls, read call transcripts, manage contacts, set and test webhooks, and check usage and billing stats. Please make it work both as a local MCP server over stdio and as an optional HTTP server with a health route and an MCP endpoint. Use an API key from environment variables, and allow auth headers in HTTP mode too.

Make the tools easy to understand, add solid error handling, and include a README with setup steps and a few natural example prompts. Look up current MCP docs online if you need to.

Want more depth? Deep Reverse