explaingit

vercel-labs/zerolang

Analysis updated 2026-06-24

4,153CAudience · researcherComplexity · 4/5Setup · moderate

TLDR

Experimental agent-first programming language from Vercel Labs. Small regular syntax, deep stdlib, and structured JSON diagnostics so AI agents can read compiler output as data.

Mindmap

mindmap
  root((zerolang))
    Inputs
      Source files
      CLI commands
      Build targets
    Outputs
      Native binaries
      JSON diagnostics
      Size reports
    Use Cases
      AI agent codegen
      Experimental language
      Reproducible builds
    Tech Stack
      C
      pnpm
      Native compiler
Click or tap to explore — scroll the page freely

Code map

Detail Auto

An interactive map of this repo's files and how they connect — its source is parsed live in your browser. Click Visualize to build it.

filefunction / class

What do people build with it?

USE CASE 1

Experiment with code generation from an AI agent against a language designed for agents.

USE CASE 2

Test how structured JSON diagnostics improve LLM repair loops compared to free-text errors.

USE CASE 3

Build a small native CLI binary for linux-musl-x64 using zero build.

USE CASE 4

Run the conformance suite to understand what the pre-1 surface currently supports.

What is it built with?

CpnpmShell

How does it compare?

vercel-labs/zerolangtorvalds/audionoisetaviso/loadlibrary
Stars4,1534,3744,483
LanguageCCC
Setup difficultymoderatemoderatehard
Complexity4/53/54/5
Audienceresearcherdeveloperresearcher

Figures from each repo's GitHub metadata at analysis time.

How do you get it running?

Difficulty · moderate Time to first run · 30min

Pre-1 and intentionally unstable. The README warns against running it on sensitive data or trusted infrastructure.

License is not stated in the available content.

In plain English

Zerolang is an experimental new programming language from a team at Vercel. The unusual thing about it is the target user. Most programming languages are designed for human developers, and then tooling for AI agents is bolted on later. Zerolang flips that around and treats AI agents as the primary users from day one, while still being usable by people. The README calls this an agent-first programming language. The project is in a very early state. The README says it is pre-1 and intentionally unstable, that breaking changes should be expected, and that today's syntax is something to explore rather than memorize. There is also a blunt warning about security: do not run it on sensitive data or trusted infrastructure, only inside an isolated, disposable environment. The goals the README lists are mostly about making the language easy for a software agent to learn and work with. The surface is meant to be small and regular, so an agent can pick it up quickly from examples and compiler feedback. The standard library is meant to be deep, so most programs do not need to go hunting for third-party dependencies. Diagnostics, dependency graph facts, size reports, and suggested fixes are returned in a structured form, which means a tool or an agent can read them as data rather than parsing free text. The language also favors one obvious way to do most things, even if that means writing more explicit code than a human might prefer. Installing it is a one-line shell command that fetches the latest release and puts a binary called zero on the path. From there the README shows a handful of commands. The zero check command type-checks a program, zero run executes it, and zero build can produce a native executable for targets such as linux-musl-x64. There are also commands like zero graph, zero size, zero skills, and zero doctor that print structured JSON output for inspection. The repository ships scripts for documentation tests, a conformance suite, native tests, command contract checks, and local benchmarks, all run through pnpm.

Copy-paste prompts

Prompt 1
Install zerolang with the one-line shell script and write a hello-world program, then run it with zero run and zero check. Show every command.
Prompt 2
Build a Claude-driven agent loop that writes a zerolang file, runs zero check, parses the JSON diagnostics, and patches the source. Sketch the full Python script.
Prompt 3
Use zero build to produce a static linux-musl-x64 binary for a small CSV-summing program, then containerize it in a 5MB Docker image.
Prompt 4
Compare zerolang vs Mojo vs Go for an LLM agent generating CLI tools in 2026. Focus on diagnostic quality and stdlib depth.
Prompt 5
Run the zero graph and zero size commands on the conformance suite, then write a Node script that parses the JSON output and prints the largest modules.

Frequently asked questions

What is zerolang?

Experimental agent-first programming language from Vercel Labs. Small regular syntax, deep stdlib, and structured JSON diagnostics so AI agents can read compiler output as data.

What language is zerolang written in?

Mainly C. The stack also includes C, pnpm, Shell.

What license does zerolang use?

License is not stated in the available content.

How hard is zerolang to set up?

Setup difficulty is rated moderate, with roughly 30min to a first successful run.

Who is zerolang for?

Mainly researcher.

Open on GitHub → Explain another repo

This repo across BitVibe Labs

Verify against the repo before relying on details.