Getting started

What is Mirri Code

Mirri Code is an AI agent that helps you carry out software development tasks — reading and modifying code, running shell commands, searching files, fetching web pages, and autonomously planning and adjusting its next steps based on feedback as it works. It ships primarily as a desktop app, with a mirri CLI for the terminal.

It fits scenarios such as:

• Writing and modifying code: implementing new features, fixing bugs, completing refactors
• Understanding a project: exploring an unfamiliar codebase and answering questions about architecture and implementation
• Automating tasks: batch-processing files, running builds and tests, chaining multiple scripts together

The CLI is written in Rust. The official install script ships a self-contained native executable that requires no pre-installed runtime; the npm channel is retired (see the migration note below).

Installation

Mirri Code's primary distribution is the desktop app; the CLI is also available via the official install script. The npm channel is retired and remains available to existing users only.

Desktop app (recommended)

Download the installer for your platform (macOS .dmg or .zip, Windows .exe, Linux .AppImage / .deb) from GitHub Releases and install it. The macOS package is signed and notarized.

Install script

Install the CLI from a terminal — no Node.js required:

• macOS / Linux:

curl -fsSL https://mirricode.com/install.sh | bash

• Windows (PowerShell):

irm https://mirricode.com/install.ps1 | iex

On Windows, install Git for Windows before first launch for the Bash tool (Unix syntax); the PowerShell tool is available out of the box. If Git Bash is in a custom location, set MIRRICODE_SHELL_PATH to the absolute path of bash.exe.

The script automatically downloads the latest release, verifies the checksum, and places the mirri executable on your PATH.

Note

The script locates GitHub Releases by v-prefixed tag. Until the first vX.Y.Z Release is published, resolving the latest version 404s — this is expected, use the desktop installer in the meantime. Pinning an old version (e.g. 1.10.0) also 404s, because old releases use scoped tags and likely have no manifest.json.

Migrating from npm

The npm packages (@mirri-ai/mirri-code, @mirri-ai/mirri-node) are retired after their last release:

• Existing installs keep working — npm installations and runs are unaffected; no new versions are published, and no update prompts will appear.
• Move to the desktop app — download the installer from GitHub Releases, or use the install script for the CLI. Configuration is stored in ~/.mirri-code/ (see Where data is stored) and carries over as-is.
• To uninstall the npm package: npm uninstall -g @mirri-ai/mirri-code.

Upgrade and uninstall

After installation, verify that the executable is ready:

mirri --version

Upgrade: re-run the install script to overwrite with the latest release (the CLI has no mirri upgrade subcommand). Desktop users download the new installer from GitHub Releases and install over the old version.

Uninstall: if you installed via the script, delete the mirri executable (default: ~/.mirri-code/bin/mirri). If you installed via npm:

npm uninstall -g @mirri-ai/mirri-code

First launch

Open the desktop app to get started. If you prefer the browser, start the local web UI from your project directory:

cd your-project
mirri web

mirri web runs the engine in the foreground and opens the web UI in your browser; press Ctrl-C to stop it. To run a single instruction without opening any UI, use mirri print:

mirri print "Take a look at this project's directory structure"

On first launch you need to configure a model provider. In settings, add a provider and its API key (for example Mirri AI Open Platform) or a custom registry, then select a default model. Providers configured in ~/.mirri-code/config.toml work the same way.

Using other AI providers

If you want to connect Anthropic, OpenAI, Google, or other providers, edit ~/.mirri-code/config.toml directly to configure the API key. See Providers and models for details. For the full reference of all config options, see Configuration files, Environment variables, and Configuration overrides.

Your first conversation

Once a model is configured, describe a task in natural language. A good starting point is to let Mirri Code familiarize itself with the project:

Take a look at this project's directory structure and briefly describe what each directory is for.

Mirri Code automatically calls file-reading, search, and other tools to browse the relevant content before responding. Read-only operations are executed automatically by default without requiring confirmation. For operations that modify files or run shell commands, it asks for your confirmation before proceeding.

You can also describe a more concrete task directly:

Add a function in src/utils that converts any string to kebab-case, and add a unit test for it.

Mirri Code plans the steps, modifies the code, runs the tests, and tells you what it did at each step.

Slash commands

Type / in the input box to list the commands available right now; the menu filters as you type. The ones you will reach for most:

Command	Description
/new	Create a new session
/compact	Compact the session history to free up tokens
/fork	Fork this session into a new one
/plan	Toggle plan mode on/off
/auto	Auto-approve safe edits/reads
/yolo	Auto-approve everything (yolo mode)
/status	View session status

Where data is stored

Mirri Code stores its local data under ~/.mirri-code/ by default — config files, session records, logs, and the update cache. To move it elsewhere, point to a new path via the MIRRICODE_HOME environment variable. For the full directory layout, see Data locations and Environment variables.

Next steps

• Interaction and input — input box operations, approval flow, Plan mode, and YOLO mode explained
• Sessions and context — resuming sessions, compressing context, exporting sessions
• Common use cases — prompt examples for typical tasks