Anthropic models in Big-AGI

Running Anthropic in Big-AGI

Add your Anthropic API key and use Claude models at Anthropic's own API rates. Big-AGI isn't a reseller: no markup and no intermediary, so Anthropic bills you directly.

• Your key, your billing. Usage is billed by Anthropic directly, nothing routes through us.
• Effort, dialed. Set reasoning effort from low up to max and let Claude think through hard problems before it answers.
• Dynamic Filtering. On newer Claude models, one toggle upgrades hosted web search and web fetch to a version that pre-filters results before they reach Claude: more accurate, fewer tokens.
• Prompt caching and tools. Repeated context gets cached to cut follow-up costs; Claude also calls tools and reads images.

Why not just Claude's app?

claude.ai is Claude and only Claude. Big-AGI runs Claude next to GPT, Gemini, Grok, and anything else you've connected, in the same conversation, which claude.ai isn't built to do.

It also opens up what the consumer app hides: reasoning effort and temperature are real API parameters you set per request, not settings we invented. Your keys stay in your browser, not inside Anthropic's account system.

Anthropic charges per token too: a light user can pay less in a month than the subscription costs, and a heavy user isn't capped, just billed for tokens used. Beam parallel runs cost more tokens than a single Claude reply, the honest price of comparing models side by side.

Your keys and your data

Turn on Direct Connection under Anthropic's Advanced settings and the browser talks to Anthropic directly, bypassing the Big-AGI server, whenever your key is client-side and Anthropic's API allows it. Your keys stay in your browser. Chats are stored locally first and sync only if you turn it on. The AI Inspector shows the exact request before it's sent: model, parameters, token counts, and a cost estimate, including what prompt caching saved you.

Anthropic in Beam

Beam runs Claude against GPT, Gemini, or any other model you've connected, in parallel, then hands you Fusions: several strategies that combine, cross-check, and synthesize the parallel answers, which beats just picking the best one. Parallel runs use more tokens than a single chat.