SpareAPI: One key for the APIs your agent needs

Hi HN, I'm Jack. I built SpareAPI, a developer-facing API layer for AI agents that need to call real web services without managing a separate account, key, auth flow, and billing setup for every source. The current site exposes 46 services / 173 endpoints: social and developer feeds, search, finance/news, voice, and AI image/video generation. There is a live Playground, an account console with keys/usage/balance, and an MCP endpoint generated from the same catalog. Public sources like Hacker News, Product Hunt, Indie Hackers, and Y Combinator can be fetched directly at the edge. Browser-walled sources are dispatched to an online operator running ClawMoney/bnbot in a real Chrome session. Two design choices: 1. The catalog is the source of truth. It drives Discover, endpoint pages, Playground forms, MCP tools/list, MCP tools/call, and the try-it proxy. 2. Write actions are excluded from the public catalog. If a command would post, like, follow, or mutate an operator's own account, it should not be sold as a generic data API. Questions I'd like feedback on: 1. If you build agents, would you rather call web tools through MCP, plain REST, or both? 2. Which platforms are still painful enough that routing through a real browser is worth the extra latency?