Introduction

What is Mori

Not a chat window, and not an IDE copilot. Mori has hands. She carries 154 native tools across files, terminal, apps, screen control, web, and documents. Anything you connect over MCP joins them. You give her a goal in plain language. She plans it, executes it end to end, and re-checks the result before reporting.

She also improves with use. Completed tasks become playbooks. Failures become lessons. Hard-won procedures become skills that wait for your approval. Her memory follows you from the desktop to Discord, Telegram, the Mori Link iPhone app, the Mori Chrome extension, and voice.

She runs on your keys. Nine model providers are built in, plus a local Ollama server that needs no key at all. The free app is the whole app.

Install

macOS needs macOS 13 or later. Download the DMG from the download page, drag Mori to Applications, and open it. The app is signed and notarized by Apple. Grant Accessibility and Screen Recording when prompted. Those two grants power computer use; see Permissions.

Windows needs Windows 10 or 11. Run the installer from the same page. SmartScreen may warn on first run; choose More info, then Run anyway. The tool layer is native on both platforms, and the toolset is identical.

Prefer the terminal? One line installs the same app:

macOS    curl -fsSL https://studiomori.ai/install.sh | sh
Windows  irm https://studiomori.ai/install.ps1 | iex

Your first five minutes

Open Setup. Paste one key from any provider you already use, or point Mori at a local Ollama server. Setup probes the real connection: a check turns green only when a live call succeeds, and a failure names its cause. One key powers everything. Then try, in order:

1  "build me a small game and preview it"   the build-run-fix cycle
2  "remember that I prefer dark mode"       a pinned, permanent memory
3  open Skills in the sidebar               the procedures she follows
4  ask a hard question, watch the rail      plan, activity, preview
5  "run a self test"                        she proves her own toolbox

While she works, the right rail shows the plan as numbered phase boxes with honest counts, plus a live activity feed and the preview. Her thinking types into view word by word and holds until the last word lands. What you watch is what she is doing.

The rule: proof, not promises

Mori treats her own replies as claims to be checked, not facts to be believed. After any turn that changed things, a local verifier re-reads your disk and probes your ports, then appends a verdict:

✓ Verified: 3/3 changes confirmed on disk.
⚠ Verification: 1/2 confirmed. Could not verify: index.html.
⚠ Honesty check: nothing was actually sent this turn. No send tool ran.

The same philosophy gates risk. Purchases stop at the pay button. Deletes ask first. Everything she does lands in a durable audit log. The full model is in Security and approvals and Proof, not promises.

How these docs are organized

The sidebar is the book, in reading order. Getting started covers install through permissions. Core concepts explains the architecture: the life of a message, thinking stages, verification. Capabilities is one chapter per system. Surfaces and Connectivity cover everywhere she lives and everything she plugs into. Reference holds the complete tool catalog, generated from the live registry, plus the changelog and /llms.txt for agents that read these docs machine-side.