You can now run any Dockerfile on Vercel.
# π³πππππππππ.ππππππ
π΅ππΎπΌ ππππππ:π·.πΈπΊ
π²πΎπΏπ . .
πππ½ ππ πππππ -π /ππππππ .
π²πΌπ³ ["/ππππππ"]
https://t.co/xOUMi4zxpD
The new Vercel Agent is in Public Beta.
βͺοΈ Chat
βͺοΈ Investigations
βͺοΈ Plans & approvals
βͺοΈ Pull requests
βͺοΈ Read-only by default
Details in the changelog β https://t.co/mB5LSKEJcT
3 months ago I started building a coding agent that runs in the cloud.
It's since written every line of code I've shipped, including itself.
Today, I'm open sourcing it. Introducing Open Agents.
In 3 months, weekly deployments on Vercel have doubled. 30% are triggered by agents (up 1000% in 6 months).
Agents are writing software that uses AI, and agents are building agents. Infrastructure must become agentic itself. That's what we're building. https://t.co/vNdTfIiX3h
The best outcome for humanity is many strong AIs competing for the top spot.
Vercel is proudly powering https://t.co/ZsS5nRfjIF and the infrastructure that made today's model release possible.
AI slop doesn't look like slop anymore. Itβs a clean diff with green CI and a convincing PR description. It looks production-ready.
That's what makes it dangerous.
Agent responsibly.
https://t.co/x3rMjLgWM6
π The Vercel support agent is now handling ~90% of customer inquiries without human intervention.
Shockingly, we've been unable to upgrade to Opus 4.6 from 4.5 without measurable performance and customer sentiment degradation.
The "best model" is a moving target. Bullish on AI Gateways and open AI SDKs π!
Today we're open sourcing the new Chat SDK, a unified TypeScript SDK for building chat bots across Slack, Microsoft Teams, Google Chat, Discord, and more. Now in public beta.
β² ~/ πππ π ππππ
Let's see how it works β
This level of resolution is only possible because we built our support agent in-house using Vercel MCP + custom tool calls to deliver deeply personalized, end-to-end solutions.
Huge props to the team @matchai + @VishalYathish
We've reached an all-time high of 87.6% autonomous resolution rate on @vercel support cases.
Best part: people truly love it. Even when the AI can't help, the overall UX is better (we auto-fill the ticket form).
Last week I had my "CEO supports day". It's now clear to me that:
1οΈβ£ Every single legitimate support case will be automated by AI. As in: the customer needs guidance.
2οΈβ£ The remainder of tickets will be triaged and prioritized by our AI PM for coding agents to work on
As far as 1οΈβ£ goes, as we continue to fine tune our agent, evals, models, and data sources, we are confident we can achieve near 100% resolution. Because we built and deployed our support agent on @vercel, we have full control over its destiny (screenshot below).
The issue is that customers sometimes reach out to report complex product issues or even defects. As in: the product itself needs improvement! The only way to help that customer is to fix our stuff π
This is how the era of autonomous software improvement begins.
A customer reporting a problem needs to ultimately result in code changes. For most companies, this flywheel doesn't exist. Support teams get overwhelmed, root causes don't get identified, documentation doesn't improve, incidents don't get filed.
In this future, every single time you reach out to us, you'll be very directly contributing to our platform getting better.
We're rolling out a new support agent at @vercel. Not only does it solve the vast majority of our tickets, but people love it π.
π¬"The AI support agent was the first truly useful agent Iβve experienced"
π¬"This AI powered support agent Vercel has implemented is incredibly impressive"
We've been noticing people asking the agent how it was built π. We'll be sharing how we built it, but importantly, why we built vs bought.
h/t @vishalyathish@matchai
https://t.co/FYzpT3pCAm is now https://t.co/ICPFB0FVbo, the AI builder for everyone.
The new v0 uses agentic AI to plan, research, build, and debug. It works in context, across multiple steps, and adapts as you go. If you have an idea, v0 can build it.