# Tarification Windsurf : chaque plan expliqué pour 2026

Windsurf (formerly Codeium) is an AI coding IDE with a free tier, a Pro plan at $15/month (500 prompt credits), and a Teams plan at $35/seat/month. If you burn through credits fast or want to skip the subscription, RunAPI gives you access to the same underlying models — Claude, GPT-4o, Gemini — at 50% off official rates.

*Updated June 23, 2026 · RunAPI Editorial*

## Combien coûte Windsurf ?

Windsurf has three main tiers. The Free plan includes limited AI completions and chat per month with no credit card required — most developers who code daily will run out within a week or two. Pro costs $15/month and gives you 500 prompt credits; Cascade can take as many internal steps as it needs per prompt without burning extra credits. Teams is $35/seat/month with admin controls and centralized billing, but note that Teams no longer pools credits across seats — each member gets their own allocation. A separate option is bringing your own API key, which lets you choose any model and pay per token — RunAPI makes this 50% cheaper than going directly to Anthropic or OpenAI.

- **Gratuit**: Limited AI completions and chat. No credit card needed. Most daily users hit the cap within a week or two.
- **Pro — 15 $/mois**: 500 prompt credits per month. Cascade steps don't cost extra. Includes access to premium models like Claude 3.5 Sonnet and GPT-4o.
- **Teams — 35 $/siège/mois**: Centralized billing and admin dashboard. Each seat gets its own credit allocation — credits are no longer pooled across the team.
- **Clé API (BYOK)**: Plug in your own API key and pay per token. Use RunAPI to cut the cost by 50% versus official provider rates. Top up only when you need it.

## Comparaison des plans Windsurf

The table below shows what each Windsurf plan includes as of mid-2026. The Free plan is low enough that developers who code daily will hit the cap within a week or two. Pro gives you 500 prompt credits per month — each user prompt counts as one credit regardless of how many steps Cascade takes internally. Teams no longer pools credits; each seat gets its own allocation.

## Comment les tarifs Windsurf se comparent à Cursor et Claude Code

Windsurf Pro at $15/month is $5 cheaper than Cursor Pro at $20/month — both give you 500 fast requests per month, which is the same basic structure. Claude Code Max at $100/month is a different category entirely: it’s a terminal-based agent, not an editor extension. If you’re burning through credits fast on Windsurf Pro, BYOK via RunAPI lets you top up at 50% off official rates instead of upgrading to a higher plan.

### Windsurf vs Cursor

Windsurf Pro is $5/month cheaper than Cursor Pro. Both support BYOK and cap you at roughly 500 fast requests per month. Cursor has a larger extension ecosystem; Windsurf’s Cascade agent handles multi-file edits without manual context selection.

### Windsurf vs Claude Code

Claude Code Max costs $100/month and runs in your terminal, not inside VS Code or JetBrains. Windsurf at $15/month is an editor-integrated tool. The two tools are not direct substitutes.

### Coût BYOK sur RunAPI

With BYOK via RunAPI, Claude 3.5 Sonnet costs roughly half what you would pay Anthropic directly. If you maxed out your Windsurf Pro credits and need more, pay-per-token is often cheaper than upgrading.


## Comment utiliser votre propre clé API avec Windsurf

Windsurf prend en charge le BYOK sur tous les plans. Vous collez une clé API dans les paramètres et choisissez le modèle à utiliser. RunAPI fournit un point de terminaison compatible OpenAI, vous définissez donc l'URL de base sur le point de terminaison de RunAPI et utilisez votre clé RunAPI. Cela fonctionne avec chaque modèle pris en charge par RunAPI — Claude, GPT-4o, Gemini, Llama et autres. Il n'y a pas de frais d'abonnement pour RunAPI ; vous ne payez que les tokens que vous utilisez, à 50 % du tarif officiel.

### Pas de verrouillage d'abonnement

Pay-per-token means you pay for what you actually use. No monthly minimum, no seat fee, no credits burning in the background.

### Mêmes modèles, moitié du coût

RunAPI facture à 50 % des prix catalogue d'Anthropic et OpenAI. Les tokens d'entrée Claude 3.5 Sonnet coûtent 1,50 $/M via RunAPI contre 3,00 $/M en direct.

### Compatible OpenAI

Définissez l'URL de base sur RunAPI et votre code ou configuration d'outil existant fonctionne sans modification. Aucun changement de SDK requis.


## Comment utiliser RunAPI avec Windsurf

1. **Créer un compte RunAPI gratuit** — Rendez-vous sur runapi.ai et inscrivez-vous. Aucune carte de crédit n'est requise pour obtenir votre première clé API.
2. **Copier votre clé API** — Dans le tableau de bord RunAPI, générez une clé API. Notez l'URL de base affichée sur la page des clés — vous en aurez besoin à l'étape suivante.
3. **Ouvrir les paramètres Windsurf** — Dans Windsurf, allez dans Settings > AI > API Keys. Sélectionnez le fournisseur correspondant au modèle que vous souhaitez (Anthropic pour Claude, OpenAI pour GPT-4o).
4. **Saisir l'URL de base et la clé RunAPI** — Collez l'URL de base RunAPI dans le champ de point de terminaison et votre clé RunAPI dans le champ de clé API. Windsurf envoie toutes les requêtes via RunAPI, qui les proxifie au fournisseur sous-jacent à 50 % de réduction.
5. **Sélectionner votre modèle et tester** — Choisissez un modèle disponible sur RunAPI. Ouvrez un chat ou déclenchez une complétion pour confirmer que la connexion fonctionne. L'utilisation apparaît dans votre tableau de bord RunAPI en quelques secondes.

## FAQ sur la tarification Windsurf

### What happened to flow action credits?

Flow action credits are gone. Windsurf used to bill separately for each internal step Cascade took — every tool call, every file read counted against your flow credit balance. Now only your user prompts count. Cascade can run as many internal steps as it needs per prompt with no extra charge.

### I was on Pro Ultimate — what do I get now?

If you were on Pro Ultimate, Windsurf rolled you into the new Pro plan ($15/month, 500 prompt credits) and gave you 1,200 free add-on credits as a one-time delta. Some users on the old plan had 3,000+ flow actions and feel this is a downgrade — the credit math is different because flow actions and prompt credits aren’t 1-to-1.

### How many requests do I actually get per month?

Pro gives you 500 prompt credits per month. At 20 prompts a day, that’s 25 days of usage — you’ll run short at the end of the month. At 10 prompts a day, 500 credits lasts the full month comfortably. Add-on packs (250 credits for $10) roll over indefinitely, so you can top up without losing what you don’t use.

### Does Teams still have shared pooled credits?

No, Teams no longer pools credits into a shared bucket. Each seat gets its own credit allocation. If your team was on Teams specifically for the shared pool, this is a significant change — several Teams users flagged this as a reason to reconsider the plan.

### Do add-on credits expire?

Add-on prompt credit packs do not expire. You buy 250 credits for $10 and they roll over indefinitely until you use them. Monthly subscription credits (the 500 that come with Pro) reset each billing cycle and do not roll over.

### Can I set a spending cap so I don't get locked out?

Yes. You can set a maximum monthly budget for auto top-up. When your credits run low, Windsurf automatically adds a credit pack so you stay unblocked instead of getting locked out mid-session. You control the cap so you never get surprised by a large bill.

### Does RunAPI work with Cursor and Claude Code as well?

Yes. RunAPI is OpenAI-compatible and works with any tool that accepts a custom API endpoint and key. Cursor, Claude Code, Codex, and other AI editors all support BYOK configuration. You use one RunAPI key and one credit balance across all your tools — no need to manage separate provider accounts.

### What models are available through RunAPI?

RunAPI supports Claude 3.5 Sonnet, Claude 3.7 Sonnet, GPT-4o, GPT-4.1, Gemini 1.5 Pro, Llama 3.1, and a range of image, video, and audio generation models. The full model list is on the RunAPI pricing page. All text models are billed at 50% of the provider's published rate.