Reallocating $100/Month Claude Code spend to Zed and OpenRouter

The pivot

It has been reported that users — the blog’s author among them, and allegedly even an AMD senior director of AI — have been hitting Claude usage limits faster than before. Frustration is the word. Why keep paying $100 a month for a hard cap when there are more flexible ways to get the same job done? The proposed switch is simple: pay $10/month for the Zed editor and top up OpenRouter with $90 in pre-paid credits. The idea is to decouple editor subscription from model consumption so credits roll over and you only pay for what you actually use.

How the setup works

Zed provides a lightweight agent harness with Agent Client Protocol (ACP) support, letting you plug in Claude, Mistral, Gemini and others directly in-editor. The appeal: faster, less laggy editing than many VS Code forks and a native agent view that shows context and rules. Zed’s built-in credits are handy but reportedly pricier per token than going through OpenRouter, which is why the author prefers routing model calls through OpenRouter — more models, more control, and in some cases larger context windows (Zed’s native Gemini 3.1 caps at ~200k tokens while OpenRouter can surface the full 1M context for compatible models).

Privacy, rollovers and trade-offs

OpenRouter accepts pre-paid credits that expire after 365 days, so idle months don’t mean wasted subscription windows — they roll over. The author also flags privacy controls: you can opt out of having inputs/outputs used to “improve the product” (you forfeit a 1% discount if you do), and enable Zero Data Retention endpoints to minimize exposure. Trade-offs exist: some models (like certain Alibaba-hosted Qwen variants) may not be available, and Zed lags behind VS Code in extension ecosystem. But for many, that’s a price worth paying for flexibility and lower incremental costs.

Why it matters

This is less about a one-off bargain and more about a growing pattern: users want modular pricing and predictable control over AI spend. Want an editor that’s nimble? Fine. Want a multi-vendor model marketplace with roll-over credits and stronger privacy guards? Also fine. The reallocation strategy on the blog is a practical how-to for anyone tired of hitting opaque limits — and a reminder that with the AI tooling market still settling, small configuration changes can deliver disproportionately better value.

Sources: braw.dev, Hacker News