Introduction to Workjet
Workjet is the open-source AI workbench for information workers. It gives every team a single, governed workspace for AI — chat, automate, build skills — all routed through your company's own AI gateway. No scattered SaaS subscriptions, no data leaking to third-party servers, no shadow AI.
Why Workjet?
Most organizations struggle with AI adoption. Individual team members sign up for different AI tools, paste sensitive data into public chatbots, and IT has zero visibility. Workjet solves this by providing a single desktop application that routes every AI request through a centrally-managed gateway, giving your organization complete control over models, costs, data loss prevention, and audit logging.
Key Concepts
Chat
Interactive, conversational AI docked at the bottom of every tab. Drag documents onto the chat panel to discuss their contents. Switch models mid-conversation. Every message flows through the gateway for DLP scanning, cost tracking, and audit logging.
Assistants
Persistent background agents that monitor events, respond to triggers, and complete tasks without manual prompting. Draft customer emails, triage support tickets, answer Slack questions — all running autonomously. Learn more in the Assistants guide.
Engines
Cron-style scheduled automations. Generate weekly pipeline reports every Monday, scan documentation for accuracy daily, digest feature requests every morning. See the Engines guide for details.
Skills
Reusable prompt-and-tool chains that your team publishes and shares. A skill bundles a system prompt, model preference, required tools, and example inputs into a one-click workflow anyone on the team can run. Browse the Marketplace to discover community skills or create your own.
Tools (MCP Connectors)
Connect Workjet to every tool your team uses via the Model Context Protocol. Fourteen first-class connectors are available out of the box — GitHub, Slack, Google Drive, PostgreSQL, and more — with a guided connection wizard for each.
Gateway
The single chokepoint through which every AI request flows. The gateway provides model routing, data loss prevention, audit logging, cost controls, and rate limiting — giving IT and compliance teams complete visibility over how AI is used across the organization.
Architecture Overview
Workjet follows a three-zone architecture: the device (desktop app), the gateway (Cloudflare edge), and the AI providers.
+-------------------+ +-------------------------+ +------------------+
| Desktop App | | AI Gateway | | AI Providers |
| (Tauri + React) | -----> | (Cloudflare Workers) | -----> | Anthropic |
| | | | | OpenAI |
| - Chat UI | | - Model Routing | | Google |
| - Assistants | | - DLP Scanning | | Ollama |
| - Engines | | - Audit Logging | | |
| - Skills | | - Cost Controls | +------------------+
| - MCP Tools | | - Rate Limiting |
+-------------------+ +-------------------------+
|
+--------+--------+
| |
+------+------+ +------+------+
| Portal | | Marketplace |
| (Web Admin) | | (Skills) |
+-------------+ +-------------+
| Layer | Technology | Purpose |
|---|---|---|
| Desktop | Tauri 2 + React 19 | Cross-platform native app |
| Frontend | TypeScript + Tailwind CSS | Type-safe, utility-first styling |
| State | Zustand | Lightweight state management |
| API | Cloudflare Workers + Hono | Serverless edge compute |
| Database | Cloudflare D1 (SQLite) | Auth, config, metadata |
| Business DB | PlanetScale (MySQL) via Hyperdrive | Scalable business data |
| Storage | Cloudflare R2 | Documents, artifacts, skills, audit |
| Cache | Cloudflare KV | Sessions, rate limits, config cache |
| Auth | Custom OAuth2 + sessions | Google SSO, API keys, cookie sessions |
| AI Gateway | Cloudflare AI Gateway | Inference proxy with logging |
Next Steps
- Install Workjet on your platform
- Follow the Quick Start to send your first message
- Learn about configuration options
- Explore the Desktop App features