The Intelligent Workspace

An open agentic intelligence platform for research, development, and everyday work. Chat, code, write, present — with AI woven into every workflow.

Download Explore Apps
Ecosystem

One Platform, Many Workspaces

Purpose-built AI applications sharing a common intelligence fabric. Each app is tailored for a specific workflow — all powered by the same agentic runtime.

ZChat ZChat

General-purpose AI chat for home and office. Conversations with context, tool use, and multi-model support — a smarter alternative to generic chatbots.

Learn more →
ZDev ZDev

AI-native development environment. Code with an agent that understands your project, runs commands, edits files, and manages git — all in one conversation.

Learn more →
ZWord ZWord

AI document workstation. Draft, edit, and format long-form documents with intelligent assistance — from research papers to business reports.

Learn more →
ZSlide ZSlide

AI-native presentation maker. Describe your talk and get polished SVG slides — no templates, no drag-and-drop, just ideas to visuals.

Learn more →
ZFlow ZFlow

Unified workspace launcher. Your command center for all ZGent apps — open any workspace, switch contexts, and orchestrate multi-agent workflows.

Learn more →
Download All Apps
For Developers & Researchers

The Agentic Intelligence Runtime

A headless server with NeoVim-style architecture. Manage LLM conversations, tool execution, terminal sessions, and workspace storage — then connect any UI over JSON-RPC 2.0.

🧩 Headless Server

zgent-server runs as a daemon. No rendering, no UI opinions — just abstract ConversationEvent notifications that frontends interpret however they like.

🔌 Multi-Frontend

TUI client (ratatui), Tauri desktop apps, browser SPAs, Telegram bots — all connect to the same server via WebSocket or stdio pipes.

🛡️ Isolated Execution

Agent code runs in Docker containers with masked /proc, disabled ICC, and egress firewalling. Read-only mounts by default.

🧬 App Profiles

Each app is a profile.toml manifest — system prompt, tools, skills, daemon process, and HTTP config. Deploy new AI apps without touching server code.

🐍 DSL Worker

SGLang-like Python DSL for programmatic LLM control — multi-step generation, fork/join branching, structured extraction, all from within a conversation.

📦 TypeScript SDK

@zgent/sdk for WebSocket clients, @zgent/ui for shared React components, @zgent/session-runtime for framework-agnostic state.

Technical Documentation