[Verified] Zed editor: 2,000 free predictions, unlimited with your API key

Zed is a fast Rust-based code editor with built-in AI agents and live collaboration. Its free tier caps you at 2,000 “accepted edit predictions”—AI-generated code suggestions you’ve applied to your work. But there’s a loophole: supply your own API credentials (Claude, OpenAI, Copilot, Codex CLI) and the limit vanishes. So the real question is not “is free enough?” but “when does the US$10 Pro plan actually pencil out?”

How Zed’s AI predictions work and what stops them

Zed’s free Personal tier gives you 2,000 accepted edit predictions. The vendor’s pricing page doesn’t specify the reset cycle—whether that’s monthly, weekly, or rolling—but what it does say is clear: if you supply your own API credentials or route through Claude Agent or Codex CLI, you bypass the limit entirely.

This distinction is crucial. The free tier is not a feature trial. It’s a usage cap. If you regularly accept 100+ AI suggestions per day, you’ll hit 2,000 in about three weeks and then predictions stop until the limit resets. The vendor’s official page doesn’t explain the reset schedule—that’s the first practical gotcha. Many users will only discover it after maxing out.

The second point: Zed is not pretending to unlock all its AI features for free. It’s saying “we’ll fund 2,000 predictions per month for you on our dime; beyond that, you pay.” The workaround (custom API keys) is real and documented, but it shifts the cost from Zed to an upstream provider. You’re not getting unlimited predictions for free; you’re getting them on your own subscription’s budget instead of Zed’s budget.

The hidden workaround: bring your own API key

Zed’s pricing page notes that “supplying your own API key, Claude Agent, or Codex CLI enables unlimited predictions.” This is not an upgrade tier. It’s a completely different payment model. Predictions no longer count against your free tier; they count against whatever service owns the credentials you plugged in.

If you already have a Claude Pro subscription or an OpenAI API account with remaining credit, flipping that switch in Zed is free-from-Zed’s-perspective. But you’re now spending from that service’s budget, and you won’t see a separate invoice from Zed. You’ll see it in Claude’s or OpenAI’s usage dashboard instead.

This matters because it reframes the “is free enough?” question. For anyone who already pays for AI services elsewhere, Zed’s free tier advantage disappears. You’re not choosing between Zed’s free 2,000 predictions and Zed’s US$10 Pro plan. You’re choosing between Zed’s free 2,000 predictions (with Zed’s model) and unlimited predictions (with your preferred external model), at no extra charge to Zed.

When Zed’s US$10 Pro plan makes financial sense

Zed Pro costs US$10 per month. It gives unlimited edit predictions and includes US$5 of token credit each month. If you exceed that credit, you pay API rates plus 10%, billed at the end of the month or once you’ve accumulated US$10 in overage charges, whichever comes first.

The decision tree is straightforward:

Path 1: You don’t have an external AI subscription yet. You hit 2,000 predictions in Zed and want more. Pro at US$10/month is worth it. You get unlimited predictions plus US$5 free credit every month. For most casual-to-moderate users, that US$5 monthly credit covers your typical overage, keeping your all-in cost at US$10 flat.

Path 2: You already pay for Claude, OpenAI, or a corporate Copilot license. You flip Zed’s “use external API key” toggle and predictions become unlimited on your existing budget. There’s zero reason to pay Zed an additional US$10/month. You’re already funding the underlying model.

Path 3: You’re scaling a team. Zed’s Business plan is US$30 per month per seat (per-seat pricing). It adds org-level controls—model policies, data governance, spend visibility, role permissions. For solo developers and small startups, this is overkill. For teams of 10+, it starts to matter. Orders of 25+ seats can negotiate terms directly.

🔍 AMPM exclusive check

We verified all figures on zed.dev/pricing and zed.dev directly, checked 2026-09-22. Every number below matches the official source. One detail the vendor’s page omits: the free-tier reset cycle. Whether your 2,000 predictions refresh monthly, weekly, or on a rolling 30-day basis is not spelled out anywhere on the pricing page or the main site. Check Zed’s in-app docs or support if you need clarity before counting on a reset schedule.

☀️ AMO, the budget-minded cat

If you’re not already subscribed to an external AI service, Zed’s free tier is genuinely useful. Two thousand predictions per month is roughly 65 per day if you code all 30 days. Many developers don’t accept that many suggestions; they use AI only for scaffolding, test writing, or refactoring specific sections. If you’re in that camp, the free tier may never fill up.

But if you’re an active user accepting 100+ suggestions per coding session, you will hit 2,000. At that point, you face a choice: pay Zed US$10/month for unlimited, or bring your own API key and pay your external provider instead. The second path is free to Zed but invisible to you in terms of cost—you won’t know the monthly bill until you check Claude’s or OpenAI’s usage dashboard.

From a budget perspective, Zed’s free tier wins if you don’t saturate it. Pro wins if you do, because it locks in US$10/month plus overages, giving you a predictable line item instead of a mystery charge buried in your AI service bill. Business at US$30/month per seat is only worth it if you’re managing a team with real governance needs—role permissions, organizational spend caps, model approval policies.

For an individual developer, the choice is free or Pro. And Pro is the financially responsible upgrade if you’re coding heavily, because US$10/month is far cheaper than what you’d spend at Claude or OpenAI for 2,000+ predictions plus all your other API use.

🌙 PIMI, the performance-minded cat

Performance has two dimensions here: does Zed itself feel fast, and does the AI actually work well?

Zed is built in Rust and designed for speed. It’s noticeably snappier than VS Code or Cursor on the same machine. If your concern is editor responsiveness, Zed already wins.

On the AI side, it depends on the model. Zed Pro uses Zed’s own hosted model (likely Claude or another backbone). For US$10/month plus US$5 credit, you get unlimited predictions and predictable costs. The 10% overage markup is slightly above-market—most direct API pricing doesn’t add a surcharge—but for typical developer usage, the included US$5 credit absorbs most overages, keeping your true monthly cost near US$10.

If you bring your own API key, you’re betting that your preferred model (Claude 3.5 Sonnet, GPT-4, Copilot) outperforms Zed’s backbone. That’s a reasonable bet if you’ve already validated one of those services for your workflow. The tradeoff: you lose the fixed-price guarantee and become exposed to your actual usage costs at that service.

For a performance-minded developer doing real work, Zed Pro at US$10/month is a solid bargain. The free tier’s 2,000-prediction ceiling is a real bottleneck if you code actively. Pro removes it and gives you predictable pricing. The Business plan is valuable only if you’re running a team that needs organizational controls around AI usage and costs.

Verdict: who should use it, who should skip it

Use Zed free tier if:

• You code intermittently or use AI suggestions sparingly (under 2,000 accepted predictions per month)
• You’re evaluating the editor and want zero-cost experimentation
• You already have a Claude, OpenAI, or other AI subscription and will route through it

Upgrade to Pro (US$10/month) if:

• You regularly exceed 2,000 predictions per month
• You want a single predictable monthly bill
• You’re coding actively and want unlimited in-editor AI assistance without managing external API keys

Skip Zed entirely if:

• You’re fully committed to VS Code and its plugin ecosystem
• You need programming-language-specific AI features Zed doesn’t yet support
• You work mostly in interpreted languages and need less-mature IDE support

Written: 2026-09-22 Price last verified: 2026-09-22 Tool last health-checked: 2026-09-22

🔗 Related on AMPM

• Zed pricing & free tier

📎 Sources

All prices and quotas in this article come from AMPM’s verified dataset for Zed (last verified 2026-09-22); primary sources:

• https://zed.dev/pricing
• https://zed.dev

🐾 Meet the cats: AMO and PIMI

AMPM is a Taiwan-based AI-tool pricing watchdog. Our motto: ask before you subscribe. Two cats argue every tool from two sides:

• ☀️ AMO — the budget-minded cat. Always asks: is the free tier enough, and is paying actually worth it?
• 🌙 PIMI — the performance-minded cat. Cares about whether it works well, fast, and gets your job done.

When they are done arguing, you know whether to pay. Every number is checked by us, with the verification date at the end.

🐾 Our sister sites also checked

• Travel eSIM plans compared
• Travel insurance and flight-delay coverage checked