Echo makes your users pay for their own AI habit

What it does

Echo is a drop-in replacement for standard AI SDKs like the Vercel AI SDK. Instead of the developer fronting API keys and praying over usage limits, end users log in once, fund a universal balance, and pay for their own tokens. The developer sets a markup percentage and keeps the spread.

The interesting bit

The economic inversion is the real hook: Echo turns inference into a user-pays utility, like a laundromat coin slot for LLMs. It wraps familiar generateText patterns so the code change is mostly swapping an import and a hook, but the business model flips from SaaS overhead to commission-based revenue.

Key highlights

• SDKs for Next.js, React, and vanilla TypeScript that wrap OpenAI, Anthropic, Gemini, and other providers.
• One OAuth login gives users a portable balance that works across any Echo-powered app.
• Developers earn revenue by marking up per-token pricing; Echo handles metering and collection.
• Starter templates cover chatbots, SPAs, and a CLI tool (though the CLI one requires manual setup from the repo).

Caveats

• The README insists you “never proxy requests,” yet the Echo Server package is explicitly an Express proxy for routing and metering LLM calls. The line between managed infrastructure and code you might have to run yourself is unclear.
• The CLI template is excluded from the standard scaffolding flow and must be installed manually from the repository.

Verdict

Worth a look if you are shipping side projects or client demos and want to monetize without building Stripe metering. Skip it if you need full control over provider routing, data residency, or if your users expect an all-you-can-eat subscription rather than a prepaid token balance.