Token Fabrics
Token Fabrics builds user-controlled, local-first AI software. Its flagship Big-AGI is an expert AI workspace designed to let people compare and fuse multiple models, avoid vendor lock-in, and keep control of their data and API keys.
At a Glance
- AI experts and researchers
- Physicians and high-stakes professional users
- Developers, founders, and technical teams
- Organizations needing self-hosted or on-premises AI
- +1 more
AI Tools by Token Fabrics
(1)
Big-AGI
Multi Model AI Chat Workspace
Discussions
No discussions yet
Be the first to start a discussion about Token Fabrics
Latest News
Big-AGI added Claude Code chat export with resumable sessions or chat-folder ZIPs and added Mistral Large 4 preview support.
Added Claude Haiku 5.5, Labs markdown display for user messages, and Beam control over shared OpenAI code sandboxes.
Added Gemini drawing/image editing, Gemini text-to-speech updates, message timestamps, and Save Chat for incognito chats.
Added GPT-6.1 Sol and an Ultrafast GPT-6 Astra service tier.
Products & Services
A local-first, multi-model AI workspace and web app for comparing several models, running Beam parallel chats, merging responses, configuring personas, using voice and attachments, inspecting requests, and connecting cloud or local models.
MIT-licensed open-source edition that users deploy and operate themselves, including via GitHub and Docker.
Hosted full workspace at no cost, with multi-model Beam, Personas, and BYOK model access.
Paid hosted subscription adding encrypted cloud backup/sync across devices, 1 GB storage, and email support from the builder.
Market Position
Token Fabrics positions Big-AGI against single-model chat products and closed, vendor-controlled AI workspaces by emphasizing multi-model Beam/Fusion, model-provider choice, BYOK economics, local-first storage, inspectable requests, and self-hosting. The product's own site contrasts it with TypingMind and the broader ChatGPT/Claude/Gemini consumer subscription model; its differentiators are parallel model comparison, transparent controls, and no markup on model usage.
Leadership
Founders
Enrico Ros
Founder and manager of Token Fabrics and creator of Big-AGI. Public profiles describe him as an AI builder with prior experience in AI tools and chips and at Qualcomm Ventures; his GitHub profile identifies him as Enrico Ros and the Big-AGI repository credits him as author.
Executive Team
Enrico Ros
Founder and CEO; Manager of Token Fabrics LLC
Builds Big-AGI and describes himself publicly as an AI builder; prior experience includes AI tools, chips, and Qualcomm Ventures.
Founding Story
The company was formed as the operating entity for Enrico Ros's effort to build AI tools that put users in control. Big-AGI began as a public GitHub project on March 18, 2023; the stated vision is local-first, transparent, open AI software with no vendor lock-in and user-owned model access.
Business Model
Revenue Model
Hosted subscriptions fund development. Big-AGI uses a BYOK model: users pay AI providers directly for model usage, while Token Fabrics charges for hosted Pro features; it says it does not meter, cap, or take a cut of model usage.
Pricing Tiers
Free MIT-licensed self-hosted edition; users pay their own infrastructure costs.
Free hosted tier with the full workspace and users' own provider keys.
Cloud backup, multi-device sync with 1 GB storage, and email support from the builder.
Teams arrange this plan by contacting Big-AGI.
Target Markets
- AI experts and researchers
- Physicians and high-stakes professional users
- Developers, founders, and technical teams
- Organizations needing self-hosted or on-premises AI
- Users of cloud model providers and local model runtimes
- AI experts and researchers comparing model answers and investigating disagreement
- Physicians and other high-stakes professionals who want cross-model checking
- Developers and founders needing precise model controls, API transparency, and BYOK access
- Science, engineering, and technical problem solving
- Organizations wanting self-hosted or on-premises AI tooling
- ER physician user identified as John, MD
- Science Olympiad parent/maker identified as Fred L.
- AI researcher and GitHub sponsor identified as Slava