mirror of
https://github.com/different-ai/openwork
synced 2026-05-09 16:52:02 +02:00
31 lines
1.1 KiB
Plaintext
31 lines
1.1 KiB
Plaintext
---
|
|
title: "What is OpenWork"
|
|
description: "Core concepts and system vocabulary"
|
|
---
|
|
|
|
OpenWork is the control surface for running agentic work with predictable sharing.
|
|
|
|
## Core terms
|
|
|
|
- **OpenWork app**: Desktop/mobile/web client experience layer.
|
|
- **OpenWork server**: API/control layer consumed by the app.
|
|
- **OpenWork worker**: A specific workspace runtime destination.
|
|
- **OpenCode Router**: Messaging bridge for Slack/Telegram with directory routing.
|
|
|
|
## Why teams use it
|
|
|
|
- Keep agent workflows portable (`.opencode/` + `opencode.json`).
|
|
- Isolate work by worker instead of mixing everything in one thread.
|
|
- Add approvals, audit visibility, and clear ownership boundaries.
|
|
- Run local-first, then move to hosted and messaging surfaces as needed.
|
|
|
|
## Runtime modes
|
|
|
|
- **Desktop host mode**: OpenWork app starts local OpenCode/OpenWork.
|
|
- **CLI host mode**: `openwrk` runs OpenCode + OpenWork server in terminal.
|
|
- **Remote client mode**: App connects to an existing OpenWork server URL + token.
|
|
|
|
## Mental model
|
|
|
|
Use one worker per outcome domain (ops, client A, client B). Then attach skills, commands, and automations to that worker.
|