Cursor vs CodeBuddy: What $20 vs $10 Actually Buys

Every "which AI coding tool should I buy" review compares plan names and monthly prices — usually copied from a blog that copied another blog. What actually decides your bill is what the subscription meters, and whether you can steer it toward cheap models. So this comparison is built from four official sources: both pricing pages, CodeBuddy's subscription doc, and Cursor's API-key doc, all checked today.

One vendor publishes its meter reading. The other publishes its tariff. That asymmetry matters more than the $10 gap.

The price tags, side by side

	Cursor	CodeBuddy
Free	$0, no credit card	$0, 100 credits/month
Personal	Pro $20/mo · Pro+ $60/mo · Ultra $200/mo	Pro $10/mo ($96/yr ≈ $8/mo)
Teams	$40/user/mo · Premium seat $120/user/mo	$40/seat/mo
Enterprise	Custom	Contact sales
Annual discount	"Save 20% with yearly billing"	Pro $96/year, per the docs FAQ
Taxes	"All prices are exclusive of any applicable taxes."	Not stated on the pricing page

Three details the table can't carry:

• Cursor scales by multiples, CodeBuddy scales by credits. Cursor's own plan cards define the paid upgrades as limits — Pro+ is "3x Pro limits on Agent," Ultra is "20x Pro limits," the Teams Premium seat is "5x Standard limits." CodeBuddy's upgrades are enumerated credits: Pro = 2,000/month, Team = 1,000/seat/month.
• Team seats cost the same $40 on both sides. What differs is what's inside a seat — covered below.
• CodeBuddy's Pro card displays a struck-through "$20/mo" next to the $10 price (as shown on the pricing page, September 15, 2026). Cursor Pro's list price is exactly $20. Make of that what you will.

What the subscription actually buys

CodeBuddy meters credits, and says how many you get. Free = 100 credits/month. Pro = 2,000 credits/month (1,000 base + 1,000 bonus), issued monthly, no rollover. Team = 1,000 base credits per seat per month, plus 1,000 bonus per seat while the current "double credits" promotion runs. Credits are the consumption unit for everything: "code autocomplete and quick Q&A consume fewer credits," while "generating complete code blocks from complex requirements or deep code analysis consumes more." Run out mid-month and the escape hatch is a top-up pack — 500 credits for $15, valid one month, non-refundable, active subscription required.

Cursor meters "usage," and doesn't say how much you get. From the pricing page's own FAQ:

"Every plan includes a set amount of model usage. On-demand usage allows you to continue using models after your included amount is consumed, billed in arrears." — Cursor Pricing, retrieved September 15, 2026

The included amounts aren't published on the pricing page — you see the overage mechanism, not the allowance. The detailed accounting lives deeper in their docs.

Free tiers: what you get without a card

Cursor Hobby is free with "no credit card required," "limited Agent requests," and access to Composer.

CodeBuddy Free is $0 with 100 credits/month, 5,000 code completions, 3 automated tasks, and auto model routing. Two promotions stack on top: a 250-credit welcome bonus on sign-up (valid 14 days), and a daily-active bonus — 30 credits/day on Free, 50 on Pro — for starting at least one conversation in the client. The end date is "announced separately," and web-only activity doesn't count.

The 7-day Pro trial is really a checkout step: it requires linking a credit card and enabling auto-renewal, includes 500 credits, and charges your selected plan ($10/month or $96/year) when it ends. Cancel inside the trial window if you're only testing.

The question that decides your bill: can you pick a cheap model?

This is where the price tags stop mattering and the architectures diverge.

CodeBuddy: model choice is a Pro feature — with a promo-shaped asterisk. The subscription doc is blunt:

"The system automatically matches an appropriate model based on task complexity... You cannot specify a single model. Pro users can freely choose from all available models." — CodeBuddy Subscription docs, page last updated August 7, 2026

Meanwhile, the current pricing page's comparison table marks "all models" as a Free-tier benefit with a promo tag. Two official pages, two readings: either Free currently includes the full model menu as a temporary promotion, or model selection is a $10/month feature. If picking the budget tier is the entire point of your subscription, budget for Pro.

We've run the budget configuration for real: a weekend rewrite of a years-old Java blog into Rust, CodeBuddy driving GLM-5.3-Flash. The budget model did the work; rework was the exception, not the rule.

Cursor: bring your own key — for chat, and only from five providers. The official docs:

"You can add your own API keys so Cursor uses your preferred AI models. This lets you send unlimited AI messages at your own cost through providers like OpenAI, Anthropic, or Google." — Cursor Docs, retrieved September 15, 2026

Three catches, all documented on that same page:

1. "Custom API keys only work with chat models. Tab completion continues using Cursor's built-in models." The autocomplete bill stays with Cursor no matter whose key you bring.
2. OpenAI keys cover "standard, non-reasoning chat models" only.
3. The supported list is OpenAI, Anthropic, Google, Azure OpenAI, and AWS Bedrock — the docs page documents no generic OpenAI-compatible endpoint, which is how most budget hosts would plug in.

Two more footnotes from the same page: Zero Data Retention doesn't apply to BYOK traffic (your provider's privacy policy governs instead), and keys are "sent to our backend with every request" — routed through Cursor's servers for prompt building, not stored, not persisted.

One report we can't verify from here: forum threads from July 2026 describe custom-key usage still counting against the plan's "Other Models" allowance, which would contradict "unlimited at your own cost" for at least some users. Dated, unconfirmed — check current behavior before you architect your billing around BYOK.

If you do go BYOK, per-token prices decide the bill — the same arithmetic as our cheapest DeepSeek API roundup and token cost calculator.

The fine print that changes the math

• CodeBuddy repriced on August 7, 2026 (UTC+8). Pro credits doubled from 1,000 to 2,000; the list price is now $10/month. Pro users who enabled auto-renewal before that date keep $9.95/month "as long as they do not cancel" — turning auto-renewal off forfeits the protected price.
• Refunds are asymmetric. CodeBuddy monthly Pro: non-refundable. Annual Pro: refundable within 7 calendar days only if zero credits have been used. Top-up packs: non-refundable. Cursor's pricing page states no refund terms.
• Mid-cycle plan changes. Under CodeBuddy's August 7 plan, direct upgrades during an active subscription are "not yet supported" — the documented path is cancel and repurchase.
• Both auto-renew by default. CodeBuddy's docs state expiry handling explicitly: cancel auto-renewal and your quota runs to the end of the cycle; after expiry, the account drops to Free with its 100 credits.
• Payment rails. Cursor self-serve: "all major credit and debit cards." CodeBuddy: credit card, debit card, WeChat Pay, prepaid card.

Which one fits you

• Frontier-model agent work with an ecosystem around it — marketplace, cloud agents, Bugbot, Grok: Cursor at $20. Cursor's own recommendation is "Pro+ for daily agent users, and Ultra for agent power users" ($60 / $200).
• Budget-model-first workflow: CodeBuddy Pro at $10, pick the cheap tier, watch 2,000 credits. Our GLM-5.3-Flash field report is the existence proof that this tier does real work.
• Trying without a card: Cursor Hobby. CodeBuddy's trial is gated behind a card plus auto-renewal.
• Teams: $40/seat either way, so decide on stack, not price. CodeBuddy seats carry SSO, per-member usage controls, and OpenAPI access; Cursor's deeper options (pooled usage, SCIM, audit logs) sit on the Enterprise line, with a $120 Premium seat if you need 5x Standard Agent limits.

Questions people actually ask

Is CodeBuddy really half the price of Cursor? For personal paid plans, yes: $10 vs $20 list. For teams, no — $40 per seat on both sides (all checked September 15, 2026).

Can the free tiers pick models? CodeBuddy's baseline Free tier auto-dispatches and "cannot specify a single model" per the subscription doc; the pricing page currently tags all-models access as a promo. Cursor Hobby includes Composer and limited Agent requests; frontier-model access is listed under paid plans.

What happens when I run out? CodeBuddy: buy a 500-credit top-up for $15 (one month, non-refundable) or wait for the next monthly issue. Cursor: on-demand usage continues, "billed in arrears."

Do either of them refund? CodeBuddy: annual within 7 days if untouched; monthly and top-ups, no. Cursor: not stated on the pricing page.

Do these prices change? Yes. CodeBuddy moved its list price and credit structure in August 2026. Everything here was verified September 15, 2026 — re-check before you buy.

Sources

Cursor — Pricing, retrieved September 15, 2026. Cursor Docs — Bring your own API key, retrieved September 15, 2026. CodeBuddy — Pricing (Intl-English), retrieved September 15, 2026. CodeBuddy Docs — Subscription, page last updated August 7, 2026, retrieved September 15, 2026.