Skip to main content

OpenWork Developer Guide

What is this about?

OpenWork is a local-first desktop app and orchestration layer for doing work with OpenCode-powered AI agents on your own files. This guide explains how OpenWork is structured, how host and client modes work, how permissions are surfaced, and where to extend it.

Checked against primary sources

This guide is based on the official OpenWork docs and the official different-ai/openwork repository, checked on June 26, 2026.

In this section​

1. What OpenWork is​

The shortest accurate description is:

OpenWork is a local-first desktop AI workbench that packages OpenCode-powered workflows into a more shareable, permissioned, and product-like experience.

The README explicitly emphasizes:

  • local-first but cloud-ready,
  • composable runtime modes,
  • OpenCode compatibility,
  • permission requests,
  • templates,
  • skills and plugin management,
  • and remote worker connectivity.

2. The mental model​

Think of OpenWork as six cooperating layers:

LayerWhat it does
desktop appmain local user-facing shell
host moderuns a local orchestration stack
client modeconnects to an existing server
OpenCode runtimeexecutes the underlying agent workflows
permission and session UIexposes requests, todos, and timelines
skills and pluginsextend capabilities in shareable ways

3. Why developers would choose it​

  • It is explicitly designed as a more productized layer over OpenCode.
  • It supports local and remote operation.
  • It exposes permissions and execution plans visibly in the UI.
  • It wants workflows to be shareable and repeatable across teams.

4. What makes it distinctive​

OpenWork is not trying to replace OpenCode. It is trying to make OpenCode-based workflows:

  • more approachable,
  • more auditable,
  • more permission-aware,
  • and easier to share.
  1. this overview,
  2. setup and quickstart,
  3. architecture and runtime,
  4. configuration and security,
  5. extension and contribution paths.

Sources​