big-AGI

Beam is the signature feature of big-AGI: fan one prompt out to several models in parallel, compare the answers side by side, then merge or vote them into a stronger final response. Around it, the MIT-licensed workspace adds AI personas, image generation, web search with citations, code execution, document handling for PDF, Word, and Excel files, an AI Inspector that shows raw requests, and access to more than 500 models from over 20 services including OpenAI, Anthropic, Google, Mistral, and DeepSeek, all through user-supplied keys with a local-first architecture that keeps chat data on the client. Local backends are supported too, with documented setups for Ollama, LocalAI, and LM Studio for fully private use. Deployment ranges from the free hosted instance at big-agi.com to one-click Vercel deploys and self-hosted Docker, typically running in minutes on any Node-capable machine with no GPU required. The open source core is free, and an optional Pro subscription adds cloud backup and multi-device sync. At 7.1k GitHub stars, it is favored by engineers who want multi-model validation instead of a single-vendor chat window.