hi hn,
i built openwork, an open-source, local-first system inspired by claude cowork.
it’s a native desktop app that runs on top of opencode (opencode.ai). it’s basically an alternative gui for opencode, which (at least until now) has been more focused on technical folks.
the original seed for openwork was simple: i have a home server, and i wanted my wife and i to be able to run privileged workflows. things like controlling home assistant, or deploying custom web apps (e.g. our customs recipe app recipes.benjaminshafii.com), legal torrents, without living in a terminal.
our initial setup was running the opencode web server directly and sharing credentials to it. that worked, but i found the web ui unreliable and very unfriendly for non-technical users.
the goal with openwork is to bring the kind of workflows i’m used to running in the cli into a gui, while keeping a very deep extensibility mindset. ideally this grows into something closer to an obsidian-style ecosystem, but for agentic work.
some core principles i had in mind:
- open by design: no black boxes, no hosted lock-in. everything runs locally or on your own servers. (models don’t run locally yet, but both opencode and openwork are built with that future in mind.) - hyper extensible: skills are installable modules via a skill/package manager, using the native opencode plugin ecosystem. - non-technical by default: plans, progress, permissions, and artifacts are surfaced in the ui, not buried in logs.
you can already try it: - there’s an unsigned dmg - or you can clone the repo, install deps, and if you already have opencode running it should work right away
it’s very alpha, lots of rough edges. i’d love feedback on what feels the roughest or most confusing.
happy to answer questions.
Looks great. I don't use opencode myself but like a lot Claude skills, agents, I have like 50 different ones for personal life, work, dating, whatever Is there integration with Claude skills?
Have you considered using or integrating with ACP? https://zed.dev/acp
We use it in our desktop app IDE (mediar.ai) and it's pretty good
What's the security boundary here - there's no mention of a VM or anything to isolate the agent from the file system?
A current concern of mine, also when using Claude Cowork, is that I don’t want files to be modified if I can’t control their versions/ if they aren’t versioned at all.
And even though it doesn’t actually get full sudo access, giving an LLM permission to edit files without being able to track exactly what it’s doing still feels risky.
Still feels a bit technical. The Claude approach is designed for "Susan in Accounting" - what thoughts do you have to reduce the technical barriers?
Great stuff and very timely. I just started getting into using opencode and while I'm hugely optimistic about its capabilities and can use it personally without too much sweat, I was left hoping for something a bit more batteries included to give to my non technical colleagues so we can collaborate together. This looks to be exactly what we were looking for so I am looking forward to giving it a spin!
Came up with the same idea this morning and worked half day to get a working version. Then I saw this post.
Will like to contribute.
Created two days ago.... Did you use Claude Code to generate the open source code? :P
I don't understand the point when opencode desktop already exists.
sounds cool! I'm non technical, I get the workflow with controling home assistant. How could I use this at work? Like how could the workflow be between marketing and devs, for example, for website work, can I change content without dealing with GitHub?
how does it compare to opencode's own gui that uses tauri?
yo this is so sick, could def incorporate this into my workflow
god!love you!!
This doesn't seem to be open source, it's currently "all rights reserved" and not under an OSI or FSF license: https://github.com/different-ai/openwork?tab=readme-ov-file#...
The word "open" in software names has completely lost its meaning.
Personally I’m skeptical and a bit dismissive of an insta- clone of commercial offerings. I ignore these things until they’re 3-6 months old and still iterating.
Announcements on their own are like whispers.