Content being expanded: tutorial with screenshots on each provider's portal publishes at public v1.0 release. Current page provides portal URLs + estimated costs + model recommendation per task.
Why “your key, your bill”
Standard SaaS cloud model with integrated AI: you pay subscription to the SaaS tool, the tool uses a single AI provider (usually OpenAI) with its own key, charges you with 200-400% markup over real cost. Difference with our model: 5 supported AI providers, you create an account at the preferred provider, get the API key, put it in the app. You pay directly to provider, see real cost, no intermediation.
Practical advantage: for a consultant using AI occasionally (10-50 queries/month), real cost is under 5 USD/month at providers — vs 20-50 EUR/month in a SaaS with integrated AI.
How to get API keys
For each provider, the process is similar: account → billing → key generation. Portal URLs:
- Anthropic Claude:
console.anthropic.com→ API Keys → Create Key. Add credit (minimum 5 USD for initial). Cost: ~3-5 USD per 1M tokens input on Claude Sonnet, ~15-20 USD on Opus. - OpenAI GPT:
platform.openai.com→ API Keys → Create new secret key. Set monthly limit for protection. Cost: ~2-5 USD per 1M tokens input on GPT-5, ~10-15 USD on o3. - Google Gemini:
aistudio.google.com→ Get API Key. Generous free tier for small queries (15 RPM, 1M tokens/day). Cost over free tier: ~1-2 USD per 1M tokens input on Gemini 2.5 Pro. - Perplexity:
perplexity.ai/settings/api→ Generate API Key. Separate API plan from UI plan. Cost: ~1-2 USD per 1M tokens, plus citation tracking included. - DeepSeek:
platform.deepseek.com→ API Keys. Lowest prices in market: ~0.10-0.20 USD per 1M tokens input. Useful for large volumes, quality comparable to flagship models.
All keys remain encrypted locally with master password set at setup. They are never transmitted to our servers.
How to add keys in the app
Settings → AI Providers → Add Provider. Paste API key, choose default model (can be changed per task). The app does an automatic test connection to verify the key. On success, the provider becomes active in the AI Co-pilot and AI Search Optimizer menus.
You can add all 5 providers simultaneously and choose per task which to use. Or you can have only one active if you prefer consistency.
How to choose the model per task
Recommendations based on internal testing on SEO tasks:
- Audit interpretation: Claude Sonnet (clear summaries, good at prioritization) or GPT-5 (fast, cheap for large volumes).
- Content drafting (meta tags, intros): Claude Sonnet or GPT-5. Both produce natural text; Claude is more conservative on tone, GPT more flexible.
- AI Search tracking: ALL 4 (Claude, GPT, Gemini, Perplexity) — we’re interested in what each responds to queries about your brand, so you query each in turn.
- Translation (RO ↔ EN): Claude Sonnet (best on RO cultural nuances) or DeepSeek (very cheap, decent quality).
- Large volumes (1000+ queries/month): DeepSeek or Gemini Free tier — minimal costs.
- Complex reasoning (analysis of complicated SEO data): Claude Opus or o3 (more expensive but more accurate on multi-variable analysis).
Real cost monitoring
The app tracks cost per request through each provider’s published API pricing. Dashboard in Settings → AI Providers → Usage shows:
- Total monthly cost per provider
- Total cumulative cost per module (Audit interpretation vs Content drafting vs AI Search tracking)
- Trend over last 6 months
- Monthly limit setting (alert + automatic suspension on overrun)
Real total cost for a typical consultant: 2-8 USD/month for all AI tasks combined. For a power user (daily audits + heavy content drafting + AI Search tracking on 50+ queries): 10-30 USD/month.
Privacy and key security
API keys are stored encrypted AES-256 in the local SQLite database with master password (set at setup). They are NEVER transmitted in HTTP requests to our servers (no servers of ours exist that process AI calls). All API calls go DIRECTLY from your app to the chosen AI provider.
For concrete privacy audit: see Privacy audit + compliance and the GDPR vs cloud tools article.