Local API sandbox for AI coding agents. Agents write Stripe/Twilio/GitHub/OpenAI code — then run it on live keys. GhostAPI gives provider-shaped APIs on 127.0.0.1, a live traffic dashboard, MCP controls, chaos mode, and secret masking — without touching prod. npx @yiaany/ghostapi start --open

Hey Product Hunt 👋 I built GhostAPI because coding agents don’t just suggest integrations anymore — they run them. That usually means real Stripe calls, real Twilio SMS, real GitHub writes, and real OpenAI spend… often before you’re ready. GhostAPI is a local API world on 127.0.0.1: • provider-shaped endpoints (Stripe, OpenAI, Twilio, GitHub, Discord, Resend…) • live dashboard for traffic + replay • MCP tools to inspect state, force responses, and turn on chaos mode • secret masking • one-command setup for Cursor / Claude / other MCP clients Try it: npx @yiaany/ghostapi start --open Site: https://ghostapi-virid.vercel.app/ GitHub: https://github.com/yiaany/ghostapi Would love brutal feedback: 1) Which provider should be highest fidelity next? 2) What’s missing for your agent loop? 3) Would you run this in CI? I’ll be here all day — ask me anything.

love the chaos mode idea, super useful for testing how my agents handle flaky providers. one thing that would help me a lot: a way to record a real api session once and then replay it as a fixture, so i can capture tricky edge cases from production-like traffic and use them deterministically in the sandbox.

honestly this looks really useful, been burned before by agents hitting real APIs in dev. one thing that'd help me out is if the dashboard could show a diff view between expected and actual request payloads when an agent writes something weird, kinda like a replay but side by side so I can spot the hallucination faster.

GhostAPI

Local internet for AI coding agents

Local internet for AI coding agents