Start with durable concepts
Work with values, functions, persistent collections, explicit state, and host effects—not a temporary teaching notation.
A small language · a serious runtime
Start simple.
Build forever.
Bring Hara to your terminal with Homebrew, or open the Playground and start experimenting right away. The same small language stays with you as your project grows.
Install Hara Homebrew · macOS and Linux
brew install hara-lang/tap/hara 9.37×vs Python · geometric mean across 8 prepared workloads
18.52×vs Babashka · geometric mean across 8 prepared workloads
1.44×vs SBCL · geometric mean across 8 prepared workloads
235,730Hoplite /hello req/s · one worker · loopback HTTP/1.1
Hara is Lisp
One shape. A great deal of range.
The operation comes first. Parentheses hold a computation, vectors hold ordered values, and maps name facts. That small reading model continues from the first expression into a complete program.
Build from first principles
Learn the parts beneath the frameworks.
Values, functions, decisions, state, events, rendering, and effects are easier to understand when each one is visible. Hara lets you meet them directly, then compose them into a real application.
Keep the running system close
Write a form, evaluate it, inspect the value, and change the program without turning every experiment into a rebuild.
Finish a complete program
Build Tic Tac Toe from a blank canvas, combining data, rules, state, pointer input, rendering, and a live frame loop.
Or watch the parts move first — this Pong rally is one small Hara program, evaluated live by the browser kernel when you press Run.
Build the first game →The language you learn is the language you keep
Start live. Fit the runtime later.
The host can change without replacing the language or its values. Begin in the browser, then choose Java, native, or web deployment when the program needs it.
Kernel mode
Java Work inside the JVM ecosystem.
Run with the Truffle runtime, integrate with Java classes and tools, and target Native Image when a self-contained deployment is useful.
- Engine
- Truffle AST / HBC4
- Deployment
- JVM or Native Image
- Best for
- Java systems and language tooling
Native Use the primary high-performance Rust runtime.
hara-rust-full prepares whole-function WebAssembly and hosts it natively for tools, services, local applications, and embedded products.
- Engine
- Whole-Wasm
- Deployment
- Native binary or library
- Best for
- Services, CLIs, and embedded products
Web Run a real Hara kernel in the browser.
Start with the interpreter core, then load VM, trace, or whole-Wasm compilation capabilities only when the application needs them.
- Small
- Interpreter core
- Full
- VM + whole-Wasm compiler
- Best for
- Learning, tools, and browser applications
Measured in the open
Fast is a result, not a tagline.
Hara publishes the code, environment, timing boundaries, raw samples, and expected checksums behind every reference result. The full table remains visible—including the runtimes that are faster.
Eight verified algorithms
Arrays, persistent data, recursion, backtracking, branching, range processing, and matrix work share expected results across runtimes.
Preparation is not execution
Startup, preparation, first invocation, warm-up, and steady-state execution are recorded separately where the runtime supports them.
Follow the evidence
Machine details, toolchain versions, source revisions, commands, raw samples, and checksums are committed alongside the report.
Try the language, not the pitch