Big AGI

This is an open source AI workspace for people who want direct control over the language models they use. Instead of a single assistant, it connects to many providers at once with the user's own API keys, including OpenAI, Anthropic, Google Gemini, Mistral, Groq, Perplexity, DeepSeek, OpenRouter, Azure and AWS Bedrock, as well as local models through Ollama, LM Studio and LocalAI. It runs in the browser and can be installed as a web app. Its signature feature, Beam, sends one prompt to several frontier models in parallel. The models answer without seeing each other, and the user can compare the responses and merge them into one answer, using agreement between models as a check against hallucinations and disagreement as a signal to dig deeper. The workspace also offers personas with memory, detailed per model controls and an interface aimed at experts such as researchers, clinicians, lawyers and developers. Pricing is published in three levels. The open source edition is free to self host. Free costs $0 per month and adds the hosted app with advanced multi model features, personas and interface. Pro costs $9 per month billed at $108 per year and adds cloud backup of personas and chats, sync across devices, email support from the developer and priority access to new features; a monthly option is also offered, with the yearly plan saving 18 percent. Business deployments with team management and SLA guarantees are quoted on request. Every plan relies on the user's own API keys, so model usage is billed by each provider. The project is developed by Token Fabrics in San Diego and describes itself as independent and not venture funded.