Vercel releases Zero: a systems programming language designed for AI Agents, 2000+ Stars

Core conclusion

On May 15, 2026, Vercel Labs open sourced a new programming language called Zero, specially designed for AI Agents. The language is implemented in C and emphasizes explicit side effects, predictable memory management, and structured compiler output. Less than 3 days after its release, Zero has received over 2,000 stars on GitHub.

Key Points

• Release time: 2026-05-15 -Affected objects: AI Agent developers, automated workflow builders, Vercel ecosystem participants
• Core change: For the first time, AI Agent has its own system-level programming language instead of relying on Python/JavaScript wrapper
• Installation method: curl -fsSL https://zerolang.ai/install.sh | bash

Background: Why does AI Agent need its own language?

The current development of AI Agent mainly relies on Python (frameworks such as LangChain and CrewAI) or JavaScript/TypeScript. Although these general-purpose languages are flexible, they suffer from three core problems:

1. Implicit side effects — Operations such as calling external APIs, reading and writing files, and network requests are not annotated at the type system level, and the AI Agent can easily trigger unexpected behavior at runtime.
2. Unpredictable memory — The garbage collection mechanism causes memory usage to fluctuate, which is unstable for Agents that require long-running and low-cost deployment.
3. bloated — the Python interpreter + dependent libraries often weigh hundreds of MB, which is not suitable for rapid deployment and edge computing scenarios.

Zero was designed with these issues in mind.

Key Impact

Four design philosophies of Zero language

1. Explicit Effects

In Zero, functions must declare what they can do. For example, printing output requires explicitly passing the World capability object:

pub fun main(world: World) -> Void raises {
    check world.out.write("hello from zero\n")
}

AI Agent ，，「 API」。

2. Web

Zero HTTP ， Web

pub fun GET(request: Request) -> Response {
    return Response.text("hello from zero")
}

3.

Zero zero graph --json、zero routes --json ，AI Agent ——「Agent 」。

4.

C ，， linux-musl-x64 。，。

Zero vs AI Agent

• Vercel Labs Zero GitHub
• Zero
• Vercel AI Agent

Vercel、Python、LangChain、AWS Lambda

Internal link guidance

• Want to automate AI Agent? Watch: AI Agent drives automated website operations: Build a fully automatic content pipeline in 30 minutes
• Real AI Agent money-making case: 18-Year-Old Built a $5,000/mo SaaS With AI Agents — Zero Hand-Written Code
• Lower cost Agent solution: Claude Code + DeepSeek V4 Building Tutorial: API Fees Dropped by 90% (15 Minutes)