Version v0.14.2 Released

The AI-First
Programming Language

AILANG makes AI-generated code cheaper to debug, replay, and fix. Explicit effects constrain what code can do. Structured traces make errors easy to localize.

Get Started Documentation Live Demos

hello.ail

-- hello.ail
module examples/hello

export func main() -> () ! {IO} =
  print("Hello, AILANG!")

How AILANG works

Humans specify intent. Agents write .ail. AILANG constrains, runs, and explains.

Agent writes .ail

Plug AILANG into Claude Code, Gemini CLI, or any agent that follows the teaching prompt.

AILANG checks

Types, effects, capabilities, and inline contracts are verified before anything runs.

Run deterministically

Same input → same output. Capabilities (`--caps IO,FS`) constrain what code is allowed to do.

Traces explain failures

Structured per-effect traces feed back to the agent so the next iteration is targeted, not guessing.

Start in 30 Seconds

One command to install, or use your AI coding agent's plugin system

Quick Install

curl -fsSL https://ailang.sunholo.com/install.sh | bash

Claude Code

/plugin marketplace add sunholo-data/ailang_bootstrap
/plugin install ailang

Gemini CLI

gemini extensions install https://github.com/sunholo-data/ailang_bootstrap.git

Built on 12 Design Axioms

Every feature derives from non-negotiable principles that make AI-generated code easier to debug, replay, and fix.
Read the full axioms

Deterministic Execution

Same input, same output, every time. Replay any execution for debugging. No hidden nondeterminism.

Axiom 1

Effect Boundaries

Side effects are explicit in types. AI cannot hallucinate network calls in pure functions.

Axiom 3 & 4

Structured Traces

See exactly what happened. Slice traces by effect type. Get specific feedback, not "it crashed."

Axiom 2

Machine-First Design

Built for AI reasoning, not human ergonomics. Decidable structure and semantic compression.

Axiom 7

Explicit Authority

No implicit access to the world. Capabilities are statically visible and constrained by budget.

Axiom 4

Pure Functional Core

Lambda calculus, pattern matching, ADTs. Composable features that never break reasoning.

Axiom 10

Loading benchmarks...

Give your AI coding agent live AILANG knowledge

One click adds https://mcp.ailang.sunholo.com/mcp/ as a remote MCP server. Your agent then has typed tools for stdlib, examples, design docs, and benchmarks — no more scraping the docs site.

Add to Claude Desktopclaude://Add to Cursorcursor://Add to Cline (VS Code)vscode://

Or copy the config manually

For harnesses without a one-click deeplink, paste this into your MCP config file (e.g. ~/.claude/claude_desktop_config.json):

{
  "mcpServers": {
    "ailang-docs": {
      "url": "https://mcp.ailang.sunholo.com/mcp/",
      "transport": "streamable-http"
    }
  }
}

See the agent-MCP guide for the full tool catalog and examples.

See It In Action

Interactive browser demos — no installation required. All running on AILANG WebAssembly.

AI + Contracts

AI extracts structured data from documents. AILANG contracts validate every field before returning — deterministic verification of stochastic AI output.

WASMZ3AI Effect

Try it

Safe Agent

Contract-verified AI tool calling. Calculator, file reader, and SQL tools wrapped in requires/ensures contracts — invalid arguments blocked before execution.

StreamingContractsTools

Try it