VibeAround v0.6.3: a web workspace for multiple coding agents

@aria_turner That is very much the long-term bet.

For me, web gives a cleaner workspace layer while keeping the desktop app simpler and more focused. It also opens the door to collaboration and mobility much more naturally, especially with VibeAround's built-in tunnel support.

@archer_louis Thanks, this is exactly the behavior I want to learn more about.

My guess is both will happen. For everyday work, many people will still have a favorite agent. But side-by-side changes the moments where exploration matters: asking different agents for approaches, having one implement while another reviews, or checking how different model/provider choices behave on the same task.

I do not expect people to use multiple agents for every task. I want switching and comparison to become cheap enough that it feels natural when it actually matters.

@ivy_foster Thanks, that was exactly the intention.

I wanted the routing and bridge layer to be reusable and easier to reason about separately from the app UI. Publishing va-ai-api-bridge as its own crate also keeps the direction honest: VibeAround is not just wrapping tools, it needs a stable substrate for models, providers, and agents to interoperate.

@zoe_sullivan Totally agree. Parallel agents are powerful, but without good inspection they can become chaotic very quickly.

Rendered diffs and live previews are meant to make multi-agent work easier to trust. The next step is better comparison and decision-making workflows around those outputs.

@bailey_carter That is exactly the framing I like.

I do not think of VibeAround as another AI chat surface. The goal is closer to an orchestration layer for coding agents, models, provider profiles, terminals, previews, and messaging workflows.

@avery_jordan1 I like the "lightweight cloud IDE" description.

The tunnel work is actually moving toward that kind of experience: local execution and local projects, but reachable through a web workspace when you need mobility. Web terminal + live preview is an early piece of that direction.

@chloe_bennett That is one of the big product questions for me.

I am trying to keep sessions first-class rather than treating them as just chat tabs. Long term, I want VibeAround to organize work by project, agent, provider profile, branch and diff state, and live preview state, so you can come back and understand what each agent was doing.

I am also being careful with shared context. I would rather make forking and resuming explicit than blend histories together invisibly, because multi-agent work gets hard to debug once provenance disappears.

@dustin_warren Exactly. I am not trying to replace Claude Code, Codex CLI, Gemini CLI, OpenCode, or any other good agent.

The value I care about is reducing switching costs between them. VibeAround should be the place where you launch, resume, inspect changes, preview output, and route models or providers through a single surface, while each agent can still keep its own strengths.

@owen_parker4 Separation is intentional for now.

I am pretty cautious about forcing agents to share context implicitly, because I do not want to disturb each model's own session history or make debugging harder. Today, agents can still share through the workspace and files.

That said, I have a promising idea here: using ACP for explicit history fork flows, so context sharing can become deliberate instead of magical. Stay tuned.

@tessa_lynch Great question. The idea is not to replace the desktop app, but to separate responsibilities better.

I want the web workspace to become a unified entry point for heterogeneous coding agents, so users do not have to keep switching between different tools and UIs. The desktop app can stay focused on local runtime, permissions, bridge, and tunnel capabilities.

Another big reason is mobility. Since VibeAround already has built-in tunnel support, the web experience can later be accessed from other devices naturally, without asking people to install another mobile app.