Kora
An AI-native OS layer written in 370k lines of Rust
Last verified:
What is Kora?
Kora is an AI-native operating system layer written in 370k lines of Rust that runs entirely locally on your Mac. It is a cloudless AI assistant built for security from the ground up, automating workflows through voice or text commands with full file access, semantic search, and proactive AI capabilities. All processing happens locally with no cloud uploads or API calls, supporting 11 different LLM providers including Anthropic, OpenAI, OpenRouter, Groq, Mistral, DeepSeek, Fireworks, Together, xAI, Zhipu, and Alibaba.
Kora is built on ChorOS, a ground-up AI-first orchest system with graph architecture and capability-based security. The interface is conversation-driven—you simply talk or type in plain English and the OS interprets, reasons, and acts. All system resources like files, processes, devices, and user sessions are modeled as nodes in a typed graph, enabling semantic reasoning by the LLM and natural queries like
Kora pricing
Pricing model: Freemium
Paid plan starting at $69/mo (pricing verified last week). One-time purchase model with no recurring API costs. No free tier mentioned. Kora is available for Mac only. The pricing includes access to all 11 LLM providers, local AI automation, semantic search, workflow automation, and full file access without cloud uploads or API costs.
Kora pros
- 100% local processing ensures complete data privacy
- One-time purchase model with no recurring API costs
- Supports 11 different LLM providers with automatic routing
- Runs entirely on user's hardware with no cloud dependencies
- Handles semantic search, file automation, and workflow building natively
- Voice or text input with natural language understanding
- No command memorization required
- 128-bit unforgeable capability tokens for security
- Instant global token revocation
- Complete audit trails for compliance
- Zero-trust security by design
- Graph-native APIs make complex operations simple
- Writing skills instead of traditional apps
- MCP integration support for external services
- Automatic load balancing across distributed AI workloads
- Context-aware operations and semantic file organization
- Linux VM for POSIX tools support
- Proactive AI assistant with pattern learning
Kora cons
- Limited to macOS platform only
- Requires Mac with Apple Silicon (M2-M5) and 24GB+ unified memory
- Beta software in active development with bugs and breaking changes
- Not recommended for production use
- Requires significant hardware investment for workstation options
- Internet connection required for cloud AI models
- 16GB minimum storage (32GB+ recommended)
- 4GB RAM minimum (8GB recommended) but 24GB+ needed for best performance
Frequently asked questions about Kora
What is Kora?
Kora is an AI-native operating system layer written in 370k lines of Rust that runs entirely on your machine. It's a cloudless AI assistant built for security from the ground up, automating workflows through voice or text commands with full file access, semantic search, and proactive AI capabilities.
Does Kora send data to the cloud?
No. All processing happens locally with no cloud uploads or API calls. Kora has a privacy-first architecture where everything runs locally on your machine with no cloud dependencies, no tracking, and no compromises.
Which LLM providers does Kora support?
Kora supports 11 different LLM providers with automatic routing: Anthropic, OpenAI, OpenRouter, Groq, Mistral, DeepSeek, Fireworks, Together, xAI, Zhipu, and Alibaba.
What are the system requirements for Kora?
Kora requires a 64-bit ARM or x86_64 processor, 4GB RAM minimum (8GB recommended, 24GB+ unified memory for Apple Silicon M2-M5), 16GB minimum storage (32GB+ recommended), and an internet connection for cloud AI models. It runs on macOS with Apple Silicon (M2-M5), Raspberry Pi 5, standard x86_64 systems, ARM64 devices, and high-end workstations.
Is Kora ready for production use?
No. Kora (ChorOS) is beta software in active development. Users should expect bugs, breaking changes, and frequent updates. It is not recommended for production use at this time.
How does ChorOS security work?
ChorOS uses capability-based access control instead of traditional permissions. It uses 128-bit unforgeable tokens for access control where every edge in the graph requires a capability token with rights embedded directly. This provides instant global revocation, complete audit trails, and principle of least authority by design.
What is the interface for ChorOS?
The interface is conversation. You forget memorizing commands, paths, or permissions. ChorOS shifts computing from learned skill to natural conversation—you simply talk, text or voice, and the OS interprets, reasons, and acts. No command memorization required.
What is Everything is a Graph in ChorOS?
All system resources—files, processes, devices, and even user sessions—are modeled as nodes in a typed graph. Edges represent relationships and carry capability tokens. This unified abstraction lets the LLM reason about system state semantically and safely, enabling natural queries like 'files I edited last week'.
What are skills in Kora?
In Kora, you write skills instead of traditional apps. Skills are reusable automation workflows that integrate with external services via MCP (Model Context Protocol). Graph-native APIs make complex operations simple, and the Linux VM provides POSIX tools support.
How do I install ChorOS?
To install ChorOS, download the beta image using: curl -O https://releases.choros.dev/choros-latest.img, then write it to your storage device with: dd if=choros-latest.img of=/dev/sdX bs=4M status=progress. First boot will guide you through configuration. See installation instructions on the website for more details.