OpenWork is a free, open-source desktop application designed to facilitate the sharing and reuse of AI workflows across multiple tools and team members. Positioned as an open-source alternative to Claude Cowork and Codex, it provides cross-platform support for macOS, Windows, and Linux. The core value proposition centers on creating AI capabilities once and deploying them across different AI agents, teammates, and machines without duplication of effort.
The application functions through an MCP (Model Context Protocol) integration that can be added to compatible AI agents including Claude Code, Cursor, Codex, and ChatGPT. This integration exposes two primary tools: search_capabilities for discovering available workflows and execute_capability for running them. Users can access OpenWork either through a dedicated desktop application or directly from their existing AI agent, making it flexible for different workflow preferences. The MCP brings assigned skills, plugins, MCP connections, and integrations with Google Workspace and Microsoft 365 into any compatible agent environment.
For organizational use, OpenWork Den serves as the control plane for managing OpenWork deployments across teams. This administrative interface enables inference provisioning at scale, member and team management with granular access controls, desktop policy configuration, local model restrictions, and app version control. Organizations can publish skills and plugins through marketplaces and assign them selectively to the organization, specific teams, or individual users. The system also supports importing Anthropic-compatible plugins and making their capabilities available through the OpenWork MCP.
Local development is supported through pnpm with specialized tooling for managing multiple git worktrees simultaneously. The development environment includes configuration options for isolated profiles, mock keychain support, and remote debugging capabilities via Chrome DevTools Protocol. The project maintains comprehensive documentation at openworklabs.com/docs and provides installation options that integrate directly with existing AI agent workflows.
