Bring your own key
Every model the pipeline calls runs on a key you connect yourself. Your account, your invoice, at your provider's published price. We charge for the software and nothing else.
Your model keys, per site. Read by the pipeline on every run.
Writing, research and scoring.
Claude models for drafting and scoring.
anthropic_api_key ••••••••••••••••••••••GPT models for text generation.
openai_api_key ••••••••••••••••••••••Mistral AI models.
mistral_api_key Paste your keyIllustrating your content.
Image generation and editing.
gemini_api_key ••••••••••••••••••••••Royalty-free stock photography.
unsplash_api_key ••••••••••••••••••••••High-quality free photos.
pexels_api_key Paste your keyWhy it matters
Most AI writing tools buy tokens wholesale and sell them back to you as credits. It's a good business. It's just not a good deal.
Bundled tools quote credits, not dollars, because the markup lives in the gap. Here the invoice comes from your provider, and the cockpit shows what each run actually cost.
Negotiated rates, committed spend, free tiers, startup credits: if you have them, you keep them. A reseller would quietly pocket the difference.
A cheaper model ships on a Tuesday, you switch on Tuesday. You are not waiting for a vendor to requalify it and reprice their plans.
Prompts and content hit your own provider account under your own terms, retention and region. Nothing is pooled through someone else’s tenancy.
Catalog
Keys live per project, so a client site can run on the client's own account. One key is enough to start. Everything else the pipeline needs, from search data to crawling, is part of the platform.
Drafting, research, outlining and scoring. Pick one per stage if you want a cheap model doing the grunt work and a sharp one writing.
Generated visuals in your style, or curated stock. Images are usually the biggest line on your provider bill, so the choice is yours.
Fine print
Encrypted, per project, and only ever used server-side by the pipeline on your runs. They are never sent to the browser, never logged in clear, and never shared between projects.
That's a fine start. One text model key gets the pipeline writing. Add an image key when you want generated visuals. Anything you don't connect simply stays off, and the search data the pipeline runs on comes with your plan either way.
Yes, and that's the point of per-project keys. Each site carries its own set, so an agency can bill compute straight to the client's account without ever touching it.
The orchestration, the search and keyword data, the crawling, the review cockpit, the topical mapping, the quality loop, the image studio, the radar, the storage, and support. Everything except the model compute, which is between you and your provider.