Self-learning AI that learns from your chats and runs on a single iMac
Open-source React components and runtimes for building AI chat
Stage
Startup
Not given
Founded
2026
Not given
Based in
GB
US
Pricing model
Freemium
Open source
Pricing
Free: £0/forever with 25K tokens/week. Pro: £15/month with 500K tokens/week. Max: £30/month with 2M tokens/week. API: £1.50 per 1M tokens, minimum £5 top-up.
The core library is free and MIT licensed. assistant-cloud offers: Free ($0, 200 MAU included), Pro ($50/mo, 500 MAU included plus $0.10 per additional MAU), Enterprise (custom MAU pricing, contact for details).
Key features
Private memory that works immediately when you teach it
Verified shared knowledge with public rejection ledger
Local MLX inference not forwarded to external API
cognira Entity for local on-device AI with optional weight retraining
OpenAI-compatible API endpoint
Served from single 16GB iMac not a GPU cluster
Background learning loop with safety filtering
Chat history, pinning, and search
One-tap Forget button to remove memories
No ads and no data sales
Streaming support for real-time chat output
Tool execution and integration
Thread persistence and history
Voice support
Generative UI rendering
Message attachments
Conversation branching
Multi-platform support (React, Vue, React Native, Ink)
Runtime adapter support (AI SDK, LangGraph, LangChain)
Fully customizable component composition
What makes it different
Self-learning from user chats with verified public ledger of what stuck
Runs on one iMac not a GPU cluster with measurable low energy use
Independent model served directly, not a wrapper around external API
Immediate private memory plus verified weight updates to shared knowledge
Works with any backend through pluggable runtime adapters
Complete source code available for customization via CLI
Production-ready runtime handling streaming, tools, and state
Free MIT license with optional paid cloud persistence
cognira
assistant-ui