ZYRAXON-AI // Documentation
Official Documentation
Complete technical reference for ZYRAXON-AI — the source-available desktop AI agent with 11 agent modes, 888 tools across 63 categories, 8,202 MCP integrations, eternal memory, self-healing and real-time screen vision.
Start
ZYRAXON Agent
Agent Modes
→All 11 agent modes, what each unlocks, and which tier holds it.
Tools Reference
→888 tools across core engine, domain tools, and control systems.
ZYRAXON Control Systems (224+)
→Real algorithms for aircraft, cars, drones, boats, rockets, and robots.
Vision System
→Real-time screen capture, frame memory, and scene analysis.
Eternal Memory
→50,000+ compressed memories that survive forever.
Self-Healing and Self-Evolution
→Zero dead ends -- the agent repairs and improves itself.
MCP Hub and Integrations
→108 curated apps, verified against their real servers.
AI Providers and Routing
→25+ providers behind one interface, routed automatically.
Permissions and Security
→A layered permission system, secret filtering, and local-first data.
ZYRAXON Provider
Endpoints
→Every endpoint in one place — 7 of them, grouped by task, none needing a key. Chat, models, speech in, speech out, and the live transcription stream.
ZYRAXON Provider Overview
→Ten branded models behind one OpenAI-compatible endpoint. The provider is closed until the repository reaches 100,000 stars — what that means, and what stays open.
The Ten ZYRAXON Models
→What each model is for, and how to pick one. Capabilities are read live, not advertised.
Sending Images, Audio, Video and PDFs
→The exact part shape for every media type, the size limits, and how to slice a long recording so it is not rejected.
Routing: How Your Input Reaches the Right Model
→Every model accepts every modality. Which upstream handles your turn is decided server-side, from live measurement — here is exactly how.
Tool Calling
→Define your own tools in OpenAI wire form and get real tool calls back, streaming or not.
Capacity, Spillover and Errors
→How the provider stays up under load, where excess traffic goes, and what each error code means.
Use It From Anywhere
→Connect Termux, OpenHands, an editor, a mobile app, an SDK or a workflow runner to the ZYRAXON line. One base URL, one model id, no adapter.
Termux on Android
→A chat box in your phone terminal: install once, run `zyraxon`, and talk to the agent. Any model, tool calling, and a full command list.
Kali Linux and NetHunter
→Kali on a laptop and Kali NetHunter on a phone both run the same client: install curl and jq, wire the environment, and pipe real tool output into the model.
Connecting OpenHands
→OpenHands takes a custom provider through a small form. The four values to enter, and how to check them before you save.
Editors and CLI Agents
→Cursor, Continue, Cline, Aider and any OpenAI-compatible editor: change the base URL and reload the model list.
Mobile Apps and Terminals
→React Native, Flutter, Kotlin and Swift clients call the endpoint directly — no SDK, no bundler, no adapter.
SDKs and Frameworks
→The OpenAI SDK, the Vercel AI SDK and LangChain, with a working tool-calling loop in TypeScript and Python.
Your Python, PowerShell or Bash Shell
→Three ready-made clients — one per interactive shell — so the provider is a function call in the shell you already have open. Any model, streaming, and the tool-calling agent loop.
Languages and Runtimes
→Go, Rust, PHP, Java and .NET call the endpoint with the standard library — no client package required.
Bots and Home Automation
→Discord, Slack, Telegram, Home Assistant and Node-RED: one HTTP call per message, no special client.
Automation and HTTP Clients
→n8n, Make, Zapier and Postman reach the endpoint through a plain HTTP request node.
Speech: Listening and Speaking
→Hear any answer aloud in its own language, and speak to the model. 322 neural voices across 75 languages, streamed as MP3 as it is written.
Speech to Text
→One endpoint that turns a recording into a transcript, or a live microphone into text as the person is still speaking. Every container, both modes, and what to do when a file is huge.
Text to Speech
→Send text, get audio. 322 neural voices across 75 languages, streamed as MP3 as it is synthesised, with the voice following the language of the text.
Multimodal Chat
→Video, audio, pictures and documents go into the same chat call as text. Every model takes every modality, and a media turn is routed to a model that can actually read it.
Features
Development
Reference
ZYRAXON vs The World
→How ZYRAXON compares with other AI coding tools.
Contributing and Community
→Fork, build, PR -- and where the community gathers.
Zyraxon Sovereign License (ZSL)
→Our source-available license -- commercial freedom, patent protection, identity safety.
FAQ and Troubleshooting
→Common questions, known issues, and the roadmap.