Xpoz solves a specific bottleneck in the AI agent stack: models are good at reasoning about social conversation but have no reliable way to reach it. Scraping is fragile, official platform APIs are expensive and rate-limited, and manual URL fetching is slow. Xpoz instead maintains its own index of posts across Twitter/X, Instagram, TikTok and Reddit, and exposes that index through a Model Context Protocol server. Add the remote URL as a connector in Claude.ai, or register it with Claude Code, Codex or Gemini CLI, and the agent can search posts, pull profiles, filter by engagement and walk hashtag threads in plain language.
Because it speaks MCP rather than a bespoke REST contract, Xpoz works with any MCP-compatible assistant, and there are also official TypeScript and Python SDKs plus a CLI for people building their own pipelines. The engineering surface is unusually complete for this category: OAuth 2.1 authentication, streamable HTTP transport, event resumption so long queries survive interruptions, smart caching with force-refresh, cursor pagination and field selection so agents do not over-fetch data.
Pricing runs on a credit currency. The Free tier allocates 500 one-time credits with a single tracked keyword — useful for evaluation but not for ongoing work, since free credits do not refresh. Pro and Max are month-to-month subscriptions with monthly credit pools and continuous monitoring slots, and Enterprise is quoted per customer. Credits are charged per query and vary by platform: Reddit and Twitter/X are cheapest, TikTok costs more and Instagram is the priciest, which is worth modelling if Instagram is central to your monitoring plan.
