I favor AntiGravity Desktop since with the Pro plan I get a lot more usage then ChatGPT and Claude offer, but there are flaws, AntiGravity is lacking features behind Codex and Claude Code A Lot! And I don’t think that’s smart, also the layout of it is not very good and can be improved, overall it has potential to beat chatgpt and claude code, but it just needs some work done on it! I trust in Google, they always do it the best in the end!
Hi @Marko_Dejanovic,
Thank you for reaching out and sharing this report. We noticed that the report contains minimal context without a description of the issue.
To help us investigate and resolve this issue accurately, could you please provide:
- Diagnostic Logs (`.txt` file) or Error Screenshot: If you encountered an error, please attach either your diagnostic logs dumped in a `.txt` file (from `Help > Toggle Developer Tools > Console` or `View > Output > Antigravity Language Server`) or an error screenshot.
- Detailed Explanation of the Issue: What specific problem or unexpected behavior did you encounter, what were you trying to do, and what happened vs. what you expected?
Hi Ambati, thanks for the response. To clarify, this is broader product feedback rather than a specific error, so I don’t have diagnostic logs to attach.
I’m referring to the overall Antigravity Desktop experience and how it currently compares with other agentic coding tools such as OpenAI Codex and Claude Code.
I really like the direction of Antigravity. The combination of an IDE, autonomous agents, terminal/browser access, parallel agents, artifacts, and verification makes it a very promising product. The generous usage on the Pro plan is also one of the main reasons I personally prefer using Antigravity.
However, I think there are several areas where the product could be substantially improved.
First, the agent workflow still feels less mature than some competing tools. I would like to see deeper agent orchestration, better subagent workflows, stronger context/session management, more powerful customization and extensibility, and more control over how agents plan, execute, verify, and iterate on tasks.
Second, reliability is important. Agents sometimes need too much manual direction, can make unnecessary changes, or can introduce regressions when modifying existing code. Better verification, smarter iteration, stronger rollback/recovery workflows, and more transparent reasoning about what changed would make the experience much more dependable.
Third, I think the UI/UX needs work. The Agent Manager, conversations, workspace/editor experience, artifacts, terminal, and other surfaces can feel somewhat disconnected. The interface could make active agents, their current tasks, progress, tool usage, changes, and pending actions much easier to understand at a glance.
I also think Antigravity could benefit from taking some of the best ideas from existing agentic coding tools rather than trying to reinvent every workflow. Things like sophisticated hooks, skills/plugins, MCP workflows, subagent delegation, context management, permissions, and isolated worktrees are particularly important for power users.
I’m not saying Antigravity is bad — quite the opposite. I think it has a very strong foundation and has the potential to become one of the best agentic development environments available. My point is that the underlying concept is excellent, but the product still needs refinement in its agent capabilities, reliability, extensibility, and UI/UX.
If useful, I can provide a detailed list of specific workflows/features that I think should be added or improved.
If you’re gonna make the post, why wouldn’t you include it? One thing EVERYONE wants to get is solutions, regardless of the position someone might have in a scenario - you’ll always be better off having a few solutions/ideas BEFORE anything else is said or done.
I’ve been building a list for making the tools, and the many different resources outside of them, far more accessible to new users and potential tutorials. WITH NO assumed knowledge! I’m new to all these tools, and everything just assumes I knew this base level and I really didn’t..Even help guides assumed the same. Google didn’t ask me for it though, I just enjoy the tools and think it would do A LOT of good