Framer AI AgentsDesign and publish professional sites with AI
hob
The independent workspace for professional agent work
26 followers
The independent workspace for professional agent work
26 followers
hob gives you one workspace for every coding agent you run. Run agents in parallel, switch models without losing context, and track every output back to the conversation that created it. Your workspace stays consistent, reliable, and predictable - organized long after the agents finish. Built for developers who direct multiple agents every day. Independent by design: bring your own keys, choose any model, and switch anytime without relying on a single AI supplier.
Free Options
Launch Team / Built With
How does this handle context handoffs when switching between models mid-session, especially if the new model has a smaller context window than the previous one?
@aslem7r For initial release we are not trying to manage context at a layer above the providers' harnesses. We are delegating to them directly as a meta harness with vastly better ux and automatic organization promises.
What the means to day is if you're switching models from the same provider - like two wildly different models via OpenCode, the underlying harness will handle that directly.
In practice, if you want to do a deliberate inter-backend handoff (which I often do), you can ask the backend directly for a summary and handoff to the other backend.
hob provides the underlying infra so every agent just "knows" how to do this in a way that's properly linked and observable (the handoff is traceable later).
We are studying inter-provider context management and a more robust alternative approach will be incorporated in the coming months, but at this moment we're not trying to out-harness the billion dollar harnesses.
Bonboarding
Love the idea of going back to previous agent sessions. I had to copy-paste context way too many times, curious to see how it works in your app. One question: if we go back to the same old sessions multiple times, does it not bloat the context of those?
But anyways, looks really cool, will give it a try!
@johnnyfekete Most of the harnesses do compaction really well today, and hob doesn't interfere with that. So even conversations that span hundreds of turns over several weeks / months stay very usable.
That said, hob tries to keep you organized around projects, workspaces and panes. Because its SOO easy to make a new pane, doing so frequently for distinct new features feels really seamless.
You get into that workflow and search and artifact linking becomes exponentially more useful - and context stays cleaner as a byproduct.
How does hob handle context when switching between models mid-conversation, and is there any latency hit from routing through your own keys versus a hosted setup?
Openfort
Great product! i feel this will become great unlock for many teams like us as we become less dependant on specific models
@joalavedra Having worked on teams filled with AI power users - we've got a whole roadmap specifically to support team collaboration in a way that makes you less dependent on any single provider.
Workflow sharing is live already, backend-agnostic conversation sharing coming this week (been alpha testing with some core users). Having the whole team in the same workspace interface, regardless of backend preferences, is already a big collaboration unlock and lets the org stay agile as the space shifts week to week.
Let us know what you think!
Finally tried hob and the parallel sessions actually feel stable, not just stacked tabs. Switching models mid-thread without losing context was the part that surprised me most.