VibeAround v0.7.2: bridge recorder, Codex Desktop proxy bypass, and smoother Windows launches
VibeAround v0.7.2 is out.
This release is mostly about making local coding-agent setups easier to debug and less fragile:
Added a Bridge recorder for local API bridge debugging. When bridge mode is enabled, you can open a live recorder from the connection popup and inspect the original request body, bridge request body, upstream response body, and bridge response body.
Improved local bridge proxy bypass for Codex launches, including local NO_PROXY coverage and Codex Desktop macOS no-proxy launch args, so system proxies like Clash are less likely to intercept 127.0.0.1 bridge traffic.
Continued Startkit and desktop-agent launch cleanup from the 0.7 line, including better Windows desktop app detection, StartApps/manual executable handling, and simpler managed toolchain behavior.
Fixed several thread/runtime and channel replay races that could affect IM-to-agent routing and callback handling.
- Filled missing Chinese UI translations around the new bridge recorder and connection UI.
A small compatibility note: if you use Codex Desktop with the local bridge, fully quit and relaunch Codex Desktop after upgrading so the new no-proxy launch args take effect.
Release notes:
https://github.com/jazzenchen/VibeAround/releases/tag/v0.7.2
Replies
@jazzenchen Nice release ... the bridge recorder feels especially useful for actually debugging agent flows instead of treating them like a black box. The proxy bypass fixes also sound like a big win for anyone running local setups with tools like Clash or mixed desktop agent.
@alexander_carter1 Yeah, agreed , especially the visibility part. Once you can trace the full request path across the bridge, it becomes way easier to understand where things break in multi agent setups instead of guessing from logs alone.
VibeAround
@alexander_carter1 @evelyn_white Totally. Logs are useful, but for multi-agent flows I often want to see the exact payload shape at each step. A small transformation or routing mismatch can look mysterious from the outside.
The recorder is basically there to make those failures less magical and more inspectable.
VibeAround
@alexander_carter1 Thanks Alexander! That was exactly the motivation. Once the bridge sits between the agent and the provider, treating it like a black box gets painful very quickly. The recorder is meant to make the full path visible: original request, transformed request, upstream response, and bridge response.
And yes, the proxy issue was one of those small local-environment problems that could waste a surprising amount of time.
WebCurate.co
The Bridge recorder sounds like a really useful addition. In my experience, debugging AI workflows is often harder than building them 😅
Being able to inspect the original request, transformed request, and responses in one place could save a lot of time when tracking down weird issues. Nice to see the focus on stability and developer experience rather than just adding more features.
VibeAround
@hosseinyazdi Thanks Hossein, I agree. AI workflow debugging often becomes harder than the initial build because the failure can sit in so many places: the app, the bridge, the model provider, the tool call, or the callback path.
This release is intentionally more about observability and stability than a shiny new feature. My hope is that it saves people time when something weird happens locally.
This looks like a very practical update. The bridge recorder is especially useful, because local agent setups can fail in weird ways and it is not always obvious whether the issue is the original request, the bridge transformation, or the upstream response.
The Codex Desktop proxy bypass also sounds important. Local 127.0.0.1 traffic getting intercepted by system proxies is exactly the kind of small environment issue that can waste a lot of debugging time.
Curious: do you see VibeAround becoming more of a debugging/control layer for multi-agent workflows, or staying focused mainly on bridge and launch stability?
VibeAround
@vlasoffchess Great question. I see VibeAround moving toward a lightweight control and debugging layer for local multi-agent workflows, while still keeping bridge and launch stability as the foundation.
The bridge/launcher pieces need to be boring and reliable first. On top of that, features like the recorder can make agent flows easier to inspect, route, and recover when something goes wrong. So the direction is not just “start agents,” but “keep them understandable and controllable while they run.”