API key setup
Connect a cloud AI provider in under two minutes. All keys are stored in macOS Keychain and never transmitted to ReWryte's servers.
How to add any key
The process is identical for all four providers:
- Open Main Window → Cloud AI.
- Find the provider card.
- Paste your key into the key input field.
- Type the model ID into the model field (see per-provider recommendations below).
- Click Save →.
After saving, the full key is never displayed again — only a masked form. Keys are stored in macOS Keychain, separate from the app's storage.
Groq — Maximum free tier
Groq runs open-source models on custom LPU (Language Processing Unit) hardware. The result: extremely fast inference and the most generous free tier of any major provider — no credit card required.
gsk_... Recommended model: llama-3.1-8b-instant - Go to console.groq.com and sign in or create a free account — no credit card required.
- Click API Keys in the left sidebar.
- Click "Create API Key", name it "ReWryte", and click Submit.
- Copy the key — it starts with gsk_.
- In ReWryte: open Main Window → Cloud AI, find the Groq card, paste the key, type
llama-3.1-8b-instantin the model field, and click Save →.
Groq free tier — full model comparison
| Model ID | Free RPD | Free RPM | Speed | Quality | Best use |
|---|---|---|---|---|---|
llama-3.1-8b-instant Our pick | 14,400 | 6,000 | Ultra-fast | Good | Daily rewrites — effectively unlimited |
gemma2-9b-it | 14,400 | 30 | Fast | Better | Quality + high volume |
mixtral-8x7b-32768 | 14,400 | 30 | Fast | Better | Longer text, nuanced tone (32K context) |
llama-3.2-3b-preview | 7,000 | 30 | Fastest | Fair | Ultra-short snippets |
llama-3.2-1b-preview | 7,000 | 30 | Fastest | Lowest | Absolute minimum latency |
llama-3.3-70b-versatile | 1,000 | 30 | Medium | Excellent | Important long documents |
RPD = Requests Per Day · RPM = Requests Per Minute
- • Day-to-day volume:
llama-3.1-8b-instant— 14,400 RPD and 6,000 RPM. You will not hit this limit in normal use. - • Better quality at high volume:
gemma2-9b-itormixtral-8x7b-32768— same 14,400 RPD, stronger output for longer or more nuanced rewrites. - • Long documents:
mixtral-8x7b-32768has a 32K context window — handles far more text per request. - • Occasional high-stakes rewrites:
llama-3.3-70b-versatile(1,000 RPD) — significantly more capable, save it for when quality truly matters. - • Watch RPM, not just RPD:
gemma2-9b-itandmixtralhave a 30 RPM cap. If doing rapid sequential rewrites, onlyllama-3.1-8b-instant's 6,000 RPM will not slow you down.
Groq hosts two separate categories of models. ReWryte calls /v1/chat/completions (the text generation endpoint). Sending that request to a prompt guard or classifier model — such as meta-llama/llama-prompt-guard-2-86m — causes an immediate 400 because those models do not support the message format and have no text generation capability.
Use one of these chat completion models instead:
| Model ID | Notes |
|---|---|
llama-3.3-70b-versatile | Best quality |
llama-3.1-8b-instant | Fastest — highest free-tier rate limits |
meta-llama/llama-4-scout-17b-16e-instruct | Good balance of speed and quality |
meta-llama/llama-4-maverick-17b-128e-instruct | Long context window |
Never use models whose names contain guard, classifier, or safety — these are not chat completion models and will always return a 400.
OpenAI
sk-proj-... Free tier: ~$5 credit on signup - Go to platform.openai.com and sign in or create a free account.
- Click your profile icon (top-right) → API keys.
- Click "+ Create new secret key", name it "ReWryte", and click Create secret key.
- Copy the key immediately — it will not be shown again after you close the dialog.
- In ReWryte: open Main Window → Cloud AI, find the OpenAI card, paste the key, type gpt-4o-mini in the model field, and click Save →.
Model options
| Model | Speed | Quality | Cost | Notes |
|---|---|---|---|---|
gpt-4o-mini Default | Fast | Good | Low | Best everyday choice |
gpt-4o | Medium | Excellent | Higher | For long or complex rewrites |
gpt-3.5-turbo | Fast | Fair | Lowest | Budget option |
Anthropic
sk-ant-... Free tier: Pay-as-you-go (no free tier) - Go to console.anthropic.com and sign in or create an account.
- Click API Keys in the left sidebar.
- Click "Create Key", name it "ReWryte", and click Create Key.
- Copy the key — it starts with sk-ant-.
- In ReWryte: open Main Window → Cloud AI, find the Anthropic card, paste the key, type claude-haiku-4-5-20251001 in the model field, and click Save →.
Model options
| Model | Speed | Quality | Cost | Notes |
|---|---|---|---|---|
claude-haiku-4-5-20251001 Default | Fast | Excellent | Low | Best for tone-following |
claude-sonnet-4-6 | Medium | Superior | Medium | Long-form rewrites |
claude-opus-4-7 | Slower | Best in class | High | Maximum quality |
Google Gemini
AIzaSy... Free tier: 1,500 requests/day free - Go to aistudio.google.com and sign in with your Google account.
- Click "Get API key" in the top-left panel.
- Click "Create API key in new project" (or select an existing project).
- Copy the generated key — it starts with AIzaSy.
- In ReWryte: open Main Window → Cloud AI, find the Google Gemini card, paste the key, type gemini-2.0-flash in the model field, and click Save →.
Model options
| Model | Speed | Quality | Cost | Notes |
|---|---|---|---|---|
gemini-2.0-flash Default | Fast | Very good | Free (1,500 RPD) | Best free choice |
gemini-1.5-flash | Fast | Good | Free (1,500 RPD) | Alternative free option |
gemini-2.0-flash-lite | Fastest | Fair | Free (1,500 RPD / 30 RPM) | Lightest, fastest |
gemini-1.5-pro | Medium | Excellent | Free (50 RPD only) | Very limited free tier |
Multi-provider strategy
Connect multiple providers for resilience. ReWryte automatically falls back to other connected providers if the active one fails or is rate-limited — no error dialogs, no manual switching.
Primary: Groq llama-3.1-8b-instant
Fastest, most free requests (14,400 RPD)
Secondary: Gemini gemini-2.0-flash
1,500 additional free RPD on a separate provider
Tertiary: Anthropic or OpenAI
Paid but highest reliability and quality ceiling
With this setup you have effectively unlimited free rewrites per day and automatic failover if any single provider is unavailable.
Key security
| What happens to your key | Detail |
|---|---|
| Stored in macOS Keychain | Separate from app storage. Even uninstalling ReWryte leaves keys in Keychain. Remove them from Cloud AI first for a clean removal. |
| Never sent to ReWryte servers | API calls go directly from your machine to the provider. There are no ReWryte proxy servers in the chain. |
| Masked in the UI after saving | The full key is never displayed again — only a masked indicator. ReWryte cannot read the key back from Keychain; it only passes it through. |
| Never written to disk in plain text | The key is stored only in the system keychain, not in any file on your machine. |
To remove a key: Main Window → Cloud AI → click the provider card → Remove (with confirmation dialog).