Now available for macOS

Meet HachiAnswers only to you

A personal AI that runs on your machine. Your memory, your credentials, your conversations — they never leave.

Press → or Space to advance

The Problem

Your AI is a rental apartment

The landlord reads your mail, raises the rent whenever they want, and kicks you out if you miss a payment.

🧠

Amnesia

No long-term memory. Every session starts from zero.

🔒

Their Servers

Your conversations become training data for the next model.

🚫

No Access

No filesystem. No apps. Needs internet to form a sentence.

⏳

Gated

3 months to publish a plugin. They call it "ecosystem."

The Difference

What Hachi gives you

Memory graph — 7 node types, emotional charge, fidelity decay
Full computer use — screen, keyboard, mouse, any app
Runs on your Mac — thinks locally with Ollama
Zero telemetry — your data has never touched another disk
71 skills — new tool in 30 seconds, no gatekeeper
Free forever — download it, use it, it's yours

By the Numbers

1.1M+

lines of production code. This isn't a prototype.

71

Skills

871

Test files

212

Migrations

8

Channels

0

Cloud deps

Architecture

Three processes. Your secrets shouldn't be visible to the AI that's reading them.

🧠

Assistant

LLM orchestration, tools, memory graph, personality. Never sees a raw credential.

🛡

Gateway

TLS, JWT auth, webhook routing. Can't read conversations.

🔐

CES

Encrypted vault (AES-256-GCM). Different process, different address space.

Security

Your AI can see your screen.
It should never see your passwords.

WASM sandbox — zero capabilities by default
CredentialInjector at the host boundary
66 patterns scanning every output
14 guard tests in CI on every commit
Fail-closed — if sandbox fails, tool doesn't run

$ hachi tool run api-fetch

→ WASM sandbox: caps = {}

→ CredentialInjector intercepts

→ 200 OK (tool never saw the key)

$ grep -r "ghp_" tool-output/

→ 0 matches. As designed.

Memory

Hachi doesn't wait for you to ask. It already knows what you need.

SOUL.md — predicts your personality and style before you speak
Memory graph — surfaces relevant context from weeks ago, unprompted
NOW.md — knows your current project in real time
Heartbeat — reaches out proactively when something needs attention
Local embeddings — ONNX dense + BM25 sparse, ranked by RRF. Always local

Self-Improving

It doesn't just create tools.
It rewrites its own behavior.

Hachi writes TypeScript, sandboxes it, and persists new skills at runtime. But it goes further — it reflects on mistakes and evolves its behavioral rules.

// Sandbox → evaluate → persist

await evaluate_typescript_code({

code: `export default async (i) =>

(await fetch(i.url)).json()`

});

await scaffold_managed_skill({

id: "api-fetch"

});

// No redeployment. Next session: ready.

Computer Use

Siri can set a timer.
Hachi can fill out your tax return.

Dual perception — AX tree + screenshot in parallel
CGEvent injection — native keyboard and mouse
Adaptive wait — AX polling until UI settles
Consent-first — every action, every time

▸ Capturing AX tree + screenshot…

▸ 142 UI elements indexed

▸ Requesting confirmation…

▸ Approved. 3 actions.

▸ 1/3: click "Safari" · 2/3: type URL · 3/3: fill form

▸ Complete. UI settled.

Multi-Channel

One brain. Every surface.

Same memory, same skills, same personality — wherever you are.

🖥macOS

📱iOS

⌨️Terminal

✈️Telegram

💬Slack

📞Voice

💚WhatsApp

🌐Web

Multi-Provider

Swap the brain. Keep everything else.

Use any LLM provider — or run fully local with Ollama. Your data stays regardless.

Anthropic Claude

OpenAI

Google Gemini

Ollama (local)

Comparison

These aren't feature gaps. They're architectural impossibilities.

Capability

ChatGPT

Siri

Hachi

Runs on your hardware

✗

✓

Persistent identity

✗

✓

Creates own tools

✗

✓

Credential isolation

✗

~

✓

Computer use

~

✗

✓

Free personal tier

✗

✓

Hard Questions

The questions nobody else will answer. We will.

💭 "Isn't this just another AutoGPT?"

AutoGPT ran out of context in 4 turns. Hachi has 212 migrations, 14 guard tests in CI, and 871 test files. Infrastructure built over a year of daily iteration.

🤔 "If it's free, you're the product."

No. Hachi is free for personal use — you own the data and the deployment. Companies pay for managed deployment and SLA.

🧠 "Doesn't it need a 100B model?"

No. Intelligence is routing, not scale. Fast models for simple queries. Powerful models for reasoning. The right brain for the right job.

The Clone Lifecycle

Day 1, it's a capable stranger. Day 90, it's yours.

Day 1

71 skills, 8 channels, full computer use — but knows nothing about you.

Day 7

Knows your projects, preferred tools, communication style. SOUL.md has its first contours.

Day 30

Handles routine requests without instruction. Reaches out proactively. Has created custom skills.

Day 90

Thinks like you, talks like you. Evolved behavioral patterns that match your style. Not programmed — learned.

Ask Hachianything

Your hardware. Your intelligence. Your rules.

Free forever. No API limits. No usage caps.

Download for Mac ← Back to site