Not another chatbot

Chatbots reset. Chalie remembers.

Every chat conversation starts from zero. Chalie builds a persistent picture of who you are — your preferences, history, and important details — across every session.

Assistants tie you to one vendor. Chalie lets you choose.

Other assistants are locked to a single AI company. Chalie works with any AI provider — local or cloud — of your choice. Switch models without losing your history.

Cloud AI owns your data. Chalie runs on your machine.

Your conversations, memories, and personal context never leave your device. No cloud database. No training on your data. Full sovereignty over your personal data.

Automation is brittle. Chalie reasons.

Automation tools move data between apps. Chalie understands what you want, loads relevant context, reasons about the best approach, and takes action — with your approval.

Why it matters

AI today is stateless. Chalie isn't.

Every conversation you have with cloud AI starts from scratch. Chalie is built differently.

	Cloud AI	Chalie
Memory	Resets every session	Remembers everything
Your data	Stored on their servers	Stays on your machine
AI model	Locked to one vendor	Any model you choose
Background work	Only when you ask	Runs tasks autonomously
Goals	No awareness	Tracks and helps proactively
Cost	Subscription	Free & open-source
Source code	Closed	Open (Apache 2.0)

Three core capabilities

Reason

Chalie reads what you're asking and picks the best way to respond — so the AI stays focused on being genuinely helpful, not just generating text.

Remember

Chalie builds a rich picture of who you are and what you care about — tracking preferences, past conversations, and important facts so every session feels like a continuation, not a fresh start.

Execute

Tools run safely and in isolation on your machine. You review and approve anything consequential before Chalie acts.

Chalie now

Talk to Chalie...

One interface. Memory that lasts. Every model.

Who it's for

People are using Chalie to think less about doing.

💻

The developer

"I have 6 projects across 3 languages. Chalie remembers the architecture decisions, the bugs I've fixed, and what I was doing last Tuesday. I just pick up where I left off."

project context background tasks local models

📚

The researcher

"I feed Chalie papers, notes, and ideas across months of work. It connects threads I'd forgotten about and surfaces relevant context when I need it — without me asking."

long-term memory document search goal tracking

🛡️

The privacy-conscious

"I wanted AI without sending my life to someone else's server. Chalie runs on my machine, my data stays in one SQLite file. Even when I use cloud APIs, my memories and context never leave my device."

local-first zero telemetry data sovereignty

Architecture

Six commitments built into Chalie.

Human-gated execution

All consequential operations require explicit approval. Autonomy is graduated, not granted.

Local-first by default

Memory, configuration, and execution state live on your machine. Your data never leaves your device, regardless of which AI provider you use.

Your choice of AI

Swap between local and cloud AI providers without losing memory or context. Chalie is independent of any single AI vendor.

Zero telemetry

No analytics. No usage reporting. No external calls except to the AI provider you configure.

Memory-native

Six memory layers on different timescales. Memory is what makes Chalie improve over time.

Composable

Extend Chalie with small tool modules written in any language. No special SDK required — just a simple interface any developer can use.

Common questions

Things people ask before trying Chalie.

Is my data sent anywhere?

No. Chalie runs entirely on your machine. All conversations, memories, and personal data are stored locally in a single SQLite file. The only outbound calls are to whichever AI provider you configure.

Which AI models work with Chalie?

Chalie works with Ollama (local), OpenAI, Anthropic, and Google Gemini. You can switch between them at any time without losing your history. The core reasoning loop needs frontier-class models (GPT-4o, Claude Sonnet, or equivalent), but lighter background tasks like OCR and drift can run on smaller local models. Different cognitive functions can use different providers.

How much does it cost?

Chalie is free and open-source (Apache 2.0). You only pay for the AI provider you choose. No subscriptions, no usage limits, no hidden fees. If you have the hardware to run large local models, your ongoing cost can be zero.

Can I use it offline?

Yes — if you run capable local models via Ollama, Chalie works offline. Memory, search, voice, and all built-in skills run on your hardware. You'll need models large enough for the reasoning loop (32B+ recommended). The only feature that requires internet is web browsing.

What does Chalie actually remember?

Everything you tell it — and things it notices. Your name, preferences, past conversations, important decisions, goals, and patterns. Memory works in layers: short-term fades naturally, but important facts and relationships persist long-term. It's designed to work like human memory — not a search engine.

What hardware do I need?

Any machine that runs Python 3.9+. macOS, Linux, or Docker. Chalie itself is lightweight — a single Python process with SQLite. The AI model is the heavy part: Chalie's cognitive architecture needs frontier-class models (GPT-4o, Claude Sonnet, Gemini Pro, or equivalent). You can use cloud APIs or run capable local models if you have the hardware.

Early Alpha

You're early. Help shape what Chalie becomes.

Chalie is in active development. The architecture is proven and everything works, but this is the beginning — your feedback shapes the direction. Early users have an outsized impact on what gets built next.

Shape the product

Every feature request and bug report from early users gets read and considered. You influence the roadmap directly.

Fast-moving

We ship fast and fix fast. Expect breaking changes between versions — that's the pace of early development.

Built in the open

Follow every decision in the build log. The code, architecture, and reasoning are all public.

Get Started Open an issue

Free forever. Apache 2.0. No account required.

Stop repeating yourself to AI.