Really, the situaton with Agent mode is so bad, it can’t solve a login with Supabase integration, i waste all 16$ credit + 20$ gifted credit and the situation still the same, redirect wrong at login and project ruined. Don’t more use V0 in my project, start in Free using 5$ with good and acceptables performance, when upgrade to Premium performance became really bad, and waste all the credits fast to resolve one single issue and sometimes lose all the project, now with Agent mode the quelity of work are really bad and also stuck in problem.
4 Likes
Goodbye V0.. I will not give you guys a second chance. Agent mode is a failure.
Sticking with Bolt/Base44/replit/lovable.. fine tune with cursor/copilot.
V0 is trash.
4 Likes
Can you please restore the old version of v0 before agent mode? I am practically begging you at this point. Or at least communicate with your community and tell us if that is a possibility or not so we can move on and find other platforms to use. I have been trying to work with agent mode since you launched it and I am going around in circles - as soon as i get it to successfully add one feature (which takes WAY too long as it is), it breaks something else. And these are simple requests,. Not only have I accomplished NOTHING on my project since agent mode was launched, I actually went backwards….there is so much broken now I don’t even know where to begin fixing things…and I don’t seem to be able to successfully restore old versions. I have wasted a week trying to get this to work for me and I am desperately hoping you will please please please give us an option to disable agent mode. You used to be the best platform for AI coding and now you are literally unusable.
4 Likes
lol the exactly same happened for me with supabase, nice to know
it causing problem when i am trying to fix a simple error, its trying to rewrite whole new app when i just want it to fix my current script only
edit:At least give us a quick toggle between Agentic mode and normal mode. I just wanted to fix a single file instead, it spun up a whole new webapp and drained all my credit in a loop. This kind of overreach makes it unusable for simple tasks. Please let us opt out or switch modes easily.
Agree, 2 of my projects got botched once the agantic mode was started. We should have an option to de-select this mode and also select preferred model and model size
1 Like
The agent mode sucks 
please revert back to the option to use GPT or other models. I’ve been on vacation the past week and just came back without noticing the change - the agent burned through $10 in credits in 2 generations and did none of the things I asked for. I have been a huge van of vercel, but this is going to make me switch 
10x the price for less effective work is not worth it
4 Likes
Yes really unhappy with the latest update.
I need to disable agent mode to continue using v0
2 Likes
I sincerely suggest that v0 team at least provide the option to use the old model. The fact is, v0 Agent Mode clearly doesn’t seem ready yet, which has caused many client projects to come to a halt. What’s worse, we are paying for Agent Mode to handle problems that it simply cannot solve. The v0 team should properly test the new model before rolling it out, instead of updating it directly and making customers test it for you—while not even offering the choice of the old model. This creates tremendous inconvenience.
2 Likes
This update is absolutely unacceptable! The quality of the code is horrendous, and the results are far below what I expected. It feels like a huge step backwards. Not to mention, the inability to test code on a new URL without an iframe modal? Seriously?
There needs to be an option to revert this update ASAP. I’ve had enough, and I’ve already canceled my subscription. This is a major disappointment, and I’m seriously hoping the developers recognize how badly this has missed the mark.
The New V0 Agent is a Complete Disaster - My Personal Experience
I’m absolutely frustrated and need to share this. The new Agent feature is a catastrophic regression that has completely ruined V0 for me.
What’s Broken
Everything I built before is screwed: All my previous projects can’t evolve properly anymore. The Agent makes terrible architectural errors and gives recommendations that make no sense. It mixes technical and functional aspects in ways that show it doesn’t understand what it’s doing.
Vibe coding is dead: What made V0 special - that smooth development experience - is completely gone. It’s like they took everything that worked and broke it.
I’ve Already Moved to Bolt
Switched to Bolt.new and there’s no comparison. It just works. My team and I wasted hundreds of hours on V0 projects that are now basically unusable.
Where’s the V0 Team?
The technical team’s lack of response is unacceptable. We loved using V0, but this silence while everything falls apart is just unprofessional.
John Smith (CTO) - what are you doing about this? At minimum, we deserve an explanation and feedback about what went wrong.
I’m Done
I’m canceling my subscription and requesting token refunds. This is a complete failure on V0’s part. It’s disappointing because I really wanted this tool to succeed.
Anyone else dealing with this mess? How are you handling the migration away from V0?
Long-time V0 user, now former subscriber
3 Likes
I previously used the GPT-5 model and found its code to be bug-ridden, often requiring manual fixes. When I tried prompting it to correct its own errors, it underperformed, consuming tokens without resolving issues. At the time, switching to the v0-1.5-md model was an easy solution, as it produced reliable code for my needs, such as building a fairly in-depth app (still in development but functioning as designed) for just 25 credits.
Now, the Agentic mode, which appears to rely on GPT-5, is producing similar buggy code that it struggles to fix. For example, I spent 50 credits on this app that still doesn’t work as intended, a stark contrast to my experience with v0-1.5-md. The forced shift to Agentic mode, without the option to select a model, has left me in a tough spot. The higher token costs and error-prone outputs make it hard to justify continuing with v0.
I’m now weighing whether to keep using v0 in hopes that Agentic mode improves or to move my codebase to a platform that offers more predictable token usage and model flexibility. Has anyone found workarounds for optimizing Agentic mode, or can Vercel consider restoring manual model selection and per-chat token tracking to address these issues? Will Vercel reimburse credits for consistent errors caused by the reliance on the Agentic model?
Hi,
Can you plesae make option to turn off that Agent mode, as design quality is pretty bad now.
1 Like
I was playing around with the Agent mode to test it with the 20usd of free credits and all of a sudden they vanished, of course i was not notified so it just started consuming my bought credits. Am i going to be refunded for this? did this happen to anyone else? And does the team listen to anyones feedbac? Everyone hated the three tier model when we all signed up with premium plans for unlimited messaging but they didnt care. Now they put the agent mode which everyone also seems to hate and they force us to burn through credits like we are some VC backed vibe coders with money to burn. Do better guys
1 Like
Yes, I was excited when I was able to manually switch on GPT-5 but it was ponderous, buggy and I quickly lost confidence in it. It is as if GPT-5 is overthinking or something. I’ve got the same dilemma as you in that I really like the v0 platform and its integration with Vercel platform, but the Agent is driving me nuts. It is simply not as intelligent as the previous models. Even if they are secretly routing the code-writing to the old model, something or other is impeding quality and speed. It feels like a backward step.