kukicha

Kukicha

AI-assisted coding.

You describe what you want. AI writes the code. Kukicha lets you read it.

Kukicha is a programming language designed to be read by humans and written by AI agents. It compiles to Go, so your programs run fast and deploy as a single binary with no dependencies.

The Workflow

You describe what you want
        ↓
AI agent writes Kukicha
        ↓
You read and approve it  ← Kukicha makes this step possible
        ↓
kukicha build → single binary
        ↓
Ship it

You don't need to know how to write Kukicha as long as you can read it and spot when something looks wrong.

See the Agent Workflow Tutorial to get started immediately.

Here's What AI-Generated Kukicha Looks Like

See if you can follow it:

import "stdlib/fetch"
import "stdlib/slice"
import "stdlib/sort"
import "stdlib/table"

type Repo
    name string as "name"
    stars int as "stargazers_count"
    language string as "language"

function main()
    repos := fetch.Get("https://api.github.com/users/golang/repos")
        |> fetch.CheckStatus()
        |> fetch.Json(list of Repo) onerr panic "fetch failed: {error}"

    popular := repos
        |> slice.Filter(r => r.stars > 100)
        |> sort.ByKey(r => r.stars)
        |> slice.Reverse()

    t := table.New("Name", "Stars", "Language")
    for repo in popular
        t |> table.AddRow(repo.name, "{repo.stars}", repo.language)
    t |> table.Print()

What just happened: Fetch repos from GitHub, keep the popular ones, sort by stars, print a formatted table. The |> pipe passes results between steps. onerr panic means "if something fails, stop and show this message." Four stdlib packages, zero boilerplate.

Quickstart

Prerequisites

Go 1.26+ is required. If you don't have Go installed, download it here.

Install Kukicha

Option A — Binary download (no Go toolchain needed after install):

Download a pre-built binary from GitHub Releases.

Option B — Install with Go:

go install github.com/duber000/kukicha/cmd/kukicha@v0.0.20
kukicha version

Your First Project

mkdir myapp && cd myapp
kukicha init          # sets up your project and stdlib

Create a file called hello.kuki:

# hello.kuki
function main()
    name := "World"
    print("Hello, {name}!")

Then run it:

kukicha run hello.kuki

Which Command Do I Use?

What Can You Build?

A Tool for Your AI Agent (MCP Server)

import "stdlib/mcp"
import "stdlib/fetch"

function getPrice(symbol string) string
    price := fetch.Get("https://api.example.com/price/{symbol}")
        |> fetch.CheckStatus()
        |> fetch.Text() onerr return "unavailable"
    return "{symbol}: {price}"

# you can use functon or func to define a function
func main()
    server := mcp.NewServer()
    server |> mcp.Tool("get_price", "Get stock price by ticker symbol", getPrice)
    server |> mcp.Serve()

Compile to a single binary and register it with Claude Desktop or any MCP-compatible agent. Add a skill declaration and run kukicha pack to generate a machine-readable manifest alongside the binary — see the Agent Workflow Tutorial.

More examples: Automation scripts, CLI tools, Concurrent health checker, REST API server, Release tooling

Beyond Readability

Kukicha compiles to Go — single binaries, goroutines, strong typing, memory safety, fast cold starts. No runtime dependencies.

But readability alone isn't enough. AI writes nearly half of all committed code, yet 45% of it contains security flaws and AI-generated code introduces 1.7x more issues. In 2026, security debt hit 82% of organizations and AI-generated code introduces 15–18% more security vulnerabilities at enterprise scale.

The Kukicha compiler catches common AI-generated vulnerabilities at build time — SQL injection, XSS, SSRF, path traversal, command injection, and open redirects while telling both you and the AI agent what safe alternative to use instead.

For Developers

If you already know Go or Python, here's how Kukicha compares:

Every Kukicha concept maps 1:1 to Go and Python — see the Quick Reference for a full translation table.

Standard Library

41+ packages, pipe-friendly, error-handled with onerr.

See the full Stdlib Reference.

Editor Support

VS Code: Search kukicha-lang in extensions, or download the .vsix from GitHub Releases. See editors/vscode/README.md.

Zed: Open Zed → zed: install dev extension → point to editors/zed/ in this repo.

Other editors: make install-lsp and configure your editor to run kukicha-lsp for .kuki files.

All editors get syntax highlighting, hover, go-to-definition, completions, and diagnostics via the LSP.

Documentation

New to Kukicha?

• Agent Workflow Tutorial — prompt AI, read and approve, ship
• Absolute Beginner Tutorial — first program, variables, functions, lists, loops

Build something real:

• Data & AI Scripting — maps, CSV, shell, LLM
• CLI Repo Explorer — types, methods, API data
• Link Shortener — HTTP servers, JSON, REST APIs
• Concurrent Health Checker — goroutines and channels

Go deeper:

• Shell Scripters Guide — for bash users
• Production Patterns — databases, validation, retry, auth
• FAQ — coming from bash, Python, or Go
• Quick Reference — Go-to-Kukicha translation table
• Stdlib Reference — all packages

Contributing

See Contributing Guide for development setup, tests, and architecture.

Status

Version: 0.0.20 — Ready for testing Go: 1.26.1+ required License: See LICENSE