AI has increased our output a ton, but something about it doesn't feel quite right. We've never made more, and never kept less. We broke this down into a few problems that we're trying to solve. First, all of this output had nowhere to live. Slack threads ate work and it became a graveyard of untraceable, unrendered HTML and markdown files. No one could tell what version was current, context didn’t travel across models or even personal agents, and editing HTMLs meant a roundtrip with an agent each time.
The second problem was repetition. Each new session needed to be taught brand, format, and audience cues to create something useful. All of this created additional overhead.
And the third problem is visibility. I would have 8 Claude Code tabs open every day, but my work feels scattered and surface level. I didn't feel good about my work like I used to. I didn't have a deep understanding of what was being done.
We built Derive to make working in the age of AI more seamless, and more fulfilling.
Derive is the home for the work that your agents produce, where every version lives at the same permanent link. You can publish HTML, markdown, or whole site from the command line or over MCP, and each revision becomes a new version at the same URL. People can comment in line on specific copy and your agent can run the whole loop over MCP (draft, publish, read comments, revise, and close out the threads). You can even make manual inline edits to live artifacts in seconds, directly in Derive. You can try this today without an account:
curl -F file=@page.html https://derive.to/v1/drafts
Contexts are my favorite piece. A Context is an agent packaged with your data and your instructions. A Context is a sandboxed working directory with permissions attached to it. So instead of being a meat machine for customer success asking analytics questions, I can build a read-only Context and give them access to it. Or sales can use Contexts to build one pagers in a repeatable predictable process. Contexts take your data and create a process that holds your guidelines so that anything your agent builds follows them.
Derive is open-source and fully self-hostable. The license is FSL. Each release converts to Apache in two years. Nothing is held back from the self-hosted build except one caveat for now that running more than one container on Node doesn’t sync live comments and cursor presence yet.
If you're interested in the hosted version, you just pay for editors, not viewers/commentors. You're covered up to 3 editors with unlimited readers and commenters. Team is $15/editor/month and agents don't take up any seats. Or you're free to self-host, and we made sure that setting up the self-hosted version is really easy to do. and self-hosting is free at any scale.
We’re around all day today so please ask any questions you have. We would love to hear how you're currently managing the AI work sprawl and what you're using to collaborate across teams and models.