OpenWork

OpenWork

An open-source desktop app for working with 50+ LLM providers locally, with team-shared skills, MCP servers, and no vendor lock-in.

🔗 Visit OpenWork
📁 AI & Machine Learning🗣️ English📅 August 29, 2026

Description

OpenWork exists for one specific worry: what happens to your AI coding setup if the vendor you're relying on changes its pricing, shuts down a feature, or you just want to try a different model without rebuilding everything? It's a desktop application, built as an open alternative to Anthropic's Claude Cowork, that lets you work with over 50 different AI models while keeping your files local-first on your own machine, and lets teams share their setups, custom skills, and integrations instead of everyone configuring their own from scratch.

Technically, OpenWork supports 50+ LLM providers (OpenAI, Anthropic, Gemini, AWS Bedrock, and others), includes browser automation capabilities, and runs local-first with an optional cloud deployment path. An MCP gateway lets organizations share skills and MCP servers org-wide, and it's compatible with existing Anthropic plugins and can migrate SKILL.md files and MCP servers from an existing Claude Cowork setup. It's fully open-source (built on OpenCode), available for macOS, Windows, and Linux, and Y Combinator-backed. The desktop app itself is free with bring-your-own API keys; team pricing starts with the first 5 seats free, then $10/seat/month including API access and an extension marketplace, with custom enterprise pricing for SSO and self-hosted infrastructure.

💬 Our review

The short version: OpenWork's main selling point is real portability — you're not locked into one AI vendor's roadmap — and being open-source with a generous free tier makes it low-risk to try, especially if you're already invested in a Claude Cowork-style workflow.

Against Claude Cowork itself, OpenWork's pitch is explicitly vendor independence: 50+ supported models versus one, and a migration path for existing SKILL.md files and MCP servers that removes most of the switching cost. Against Cursor, which is more polished but proprietary and single-vendor by design, OpenWork trades some UI refinement for openness and auditability — you can read and fork the entire codebase. The honest caveat is that bring-your-own-API-keys means the app itself being free doesn't make usage free; you're still paying each model provider directly, and as a young open-source project it will have rougher edges than mature commercial IDEs. Pick OpenWork if vendor lock-in genuinely worries you; pick a single-vendor tool if you've already committed and just want the smoothest experience.

💰 Pricing

FreemiumApp free (BYO API keys); Team $10/seat/mo after first 5 free; Enterprise custom.

📊 Global score

45Average
🌐Availability15/100Faible

1 language · 0 platform

📄Profile75/100Bien

Profile completeness

🤖 AI-enriched data

💰 Pricing model
🆓 freemium

Desktop app free (bring your own API keys). Team: first 5 seats free, then $10/seat/month. Enterprise: custom pricing.

👥 Target audienceIndividual developers, teams using AI agents, and enterprises seeking vendor-independent AI infrastructure
🗣️ LanguagesEnglish
🌍 Target countriesGlobal
👍

Pros

50+ LLM providers supported, no vendor lock-in

Fully open-source and auditable

Local-first with optional cloud deployment

Migration path from existing Claude Cowork setups (SKILL.md, MCP servers)

First 5 team seats free

👎

Cons

Bring-your-own-API-keys means model usage costs are separate and can add up

Young open-source project — less polished than established commercial IDEs

Scheduled tasks still in Alpha

Requires comfort configuring multiple model providers

❓ Frequently asked questions

What is OpenWork?
Who is OpenWork for?
How much does OpenWork cost?
Can I migrate from Claude Cowork?
Is it worth switching from Claude Cowork?
Which tool should you pick for your case?