The programming language
for agents
zerolang explores what a programming language can look like when agents are primary users from day one. The aim is a language that is easy to learn on the fly, deterministic to inspect and repair, standard-library first, and explicit enough that most tasks have one obvious path.
$ curl -fsSL https://zerolang.ai/install.sh | bashThe current toolchain is useful for exploration, but today's syntax and APIs are not a contract. Expect breaking changes while zerolang searches for what works best for agents. Run it in a safe environment, not against production systems.
Small surface area
zerolang is aiming for a language an agent can learn while working: regular syntax, few special cases, and compiler feedback that points toward the next edit.
Fewer dependency searches
The long-term goal is a standard library broad and consistent enough that most programs start with documented APIs, not package selection.
Deterministic repair loops
The toolchain is intended to expose diagnostics, graphs, size reports, explanations, and repair plans as structured data agents can consume.
fn answer i32
ret + 40 2
pub fn main Void world World !
if == answer() 42
check world.out.write "math works\n"$ zero check examples/hello.0
hello.0:1:4 PAR100: expected '{' before block
explain: zero explain PAR100Direction
Regularity over cleverness.
zerolang favors explicit capabilities and standard-library APIs over syntax for every convenience. Some code may be more verbose for humans if that makes it easier for agents to generate, inspect, and repair.
Pre-1 by design
Today's syntax and APIs are not a contract. Breaking changes are expected while zerolang searches for what works best for agents.
Safe environments only
Security vulnerabilities should be expected. Run and develop zerolang in isolated environments, not production systems or sensitive infrastructure.
Exploration over mastery
Try the current shape, inspect the output, and send feedback. The details will move as the experiment learns.
One obvious path
The language should favor a small set of regular patterns over many interchangeable styles.
Standard library over sugar
New capability should usually live in documented APIs before it becomes new syntax.
Agent-readable tooling
Diagnostics, graph facts, size reports, and repair metadata should be available as structured output.
Explicit effects
Outside-world access, fallibility, and resource use should stay visible to both readers and tools.
No legacy promises
When a clearer agent-facing design wins, zerolang can replace old behavior instead of carrying compatibility paths forward.
DX as a goal
Checking, inspecting, explaining, and repairing code should feel direct even when the language is intentionally explicit.
Explore with us.
Install the compiler, run an example, and inspect what the experiment can do today. The most useful feedback is what helps agents work with less guesswork.