Connecting AI Providers

Sciwand works with your own AI provider accounts. You bring your own API key (or run a local model), paste it into Settings, and Sciwand uses it for chat and semantic search. Your keys are stored locally on your device and are never sent to Sciwand servers.

New to this? The sections below explain the essentials in plain language.

What is an API key?

An API key is like a password that lets Sciwand use an AI provider (such as OpenAI or Google) on your behalf. You create it once on the provider website, copy it, and paste it into Sciwand. After that, Sciwand can send your questions to that provider and bring the answers back.

• It is personal - treat it like a password and do not share it.
• It is separate from any subscription. A consumer plan (for example ChatGPT Plus or Claude Pro) does not include API access; instead you pay a small amount per use, which is often cheaper.
• Your key is stored on your device and is never sent to Sciwand servers.

The two kinds of models

• Chat models answer questions, summarize papers, and help you write. You can enable several and switch between them. See Connect a Chat Model.
• Embedding models power search across your library and the MAX (full-context) features. Exactly one is active at a time. See Connect an Embedding Model.

Supported chat providers: OpenAI (ChatGPT), Claude, Gemini, OpenRouter, DeepSeek, Groq Cloud, Grok, Together, Qwen, Perplexity, plus local options (LM Studio, Ollama) and any OpenAI-compatible Custom Endpoint. Embeddings: OpenAI, Gemini, Voyage AI, Together, plus local and custom options.

Which model should I pick?

You do not need the most expensive model - match it to the job:

• Cheapest and fastest: a small model (providers usually label these mini, flash, lite, or haiku). Great for everyday questions and summaries.
• Best accuracy: a flagship (largest) model, for hard reasoning and careful writing.
• Free: a provider with a free tier (such as Gemini or Groq), or a fully private local model. See Free and low-cost models.
• Compare many at once: OpenRouter gives one key and hundreds of models.

A simple plan: use a cheap, fast model for everyday work, and switch to a stronger one when a question is hard.

Price, speed, accuracy at a glance

• Price: small models cost the least; local models are free.
• Speed: smaller models, and anything on Groq, are fastest.
• Accuracy: the largest models are strongest for difficult work.

Do I need a paid subscription?

No. Sciwand does not require any subscription, and neither do the providers. You connect your own account and pay only for what you use (pay as you go).

One thing that surprises many people: a consumer subscription is not the same as API access. A plan like ChatGPT Plus, Claude Pro, or Gemini Advanced does not include API usage - the API is billed separately. You create an API key in the provider developer console and add a small amount of credit there. For typical research use this often costs only a few cents to a few dollars, which can be much fairer than a fixed monthly subscription.

Adding funds to your provider account

Most cloud providers need a small prepaid balance (or a card on file) before the API will respond. If Sciwand shows an error such as "quota exceeded", "insufficient balance", or "billing required", open your provider billing page and add a few dollars of credit. Several providers include a free tier - see Free and low-cost models.

Local providers (LM Studio, Ollama) are completely free: no key, no subscription, no funds - everything runs on your own computer.