Install MirrorNeuron

Use this guide to install a local MirrorNeuron runtime and verify that its CLI and services are reachable. Use the deployment installer for a supported runtime; use local mode only when you are working from a workspace checkout.

Before you begin

• Use macOS, Linux, or Windows with WSL2.
• Install and start Docker.
• Install git.
• For editable local development, also install Python 3.11+ and Elixir/Erlang.

Docker Model Runner is required only for blueprints that use local models. Docker Desktop supplies it on macOS and Windows; install the Docker Model Runner plugin separately on Linux when you need local model execution.

Easies way is using installer for our released runtime:

curl -fsSL https://mirrorneuron.io/install.sh | bash

Install the dev version from github

If you want to install from our legacy or unpublised dev versions, using the github repository:

git clone https://github.com/MirrorNeuronLab/mn-deploy
cd mn-deploy
./install.sh --help
./install.sh

The installer runs non-interactively with its default selections. Use --interactive to confirm choices or --version <release-tag> to install a matching released component set.

Warning: piping an installer into a shell executes the downloaded script immediately. Prefer the checked-out installer when you need to inspect or customize the installation.

Verify the installation

Confirm that the CLI is on your PATH and that the runtime can start:

mn --help
mn runtime start
mn runtime health
mn runtime doctor

mn --help should display the MirrorNeuron CLI command groups. mn runtime health should complete without reporting a failed required runtime component. mn runtime doctor should prob all components without problem.

The default local state directory is ~/.mn. Unless you override MN_API_PORT, the REST API is available at http://localhost:54001/api/v1.

Prepare local models only when required

You usually do not need to install models manually. MirrorNeuron installs the required model automatically when a blueprint needs it. To verify that a model works on your machine, run mn doctor. It reports model availability and shows whether inference is using available hardware acceleration, including NVIDIA, AMD, Intel, or Apple Metal.

mn model list
mn model install gemma4:e2b
mn model doctor gemma4:e2b

mn model doctor checks model availability and compatibility. See Model Runtime for model configuration and remote-provider options.

Troubleshoot installation

• Symptom: mn: command not found
  • Check: which mn
  • Resolution: rerun the installer, then open a new shell so its PATH changes apply.
• Symptom: the runtime does not start
  • Check: docker info and mn runtime health
  • Resolution: start Docker, then rerun mn runtime start.
• Symptom: blueprint validation reports a missing model
  • Check: mn model doctor <model-id>
  • Resolution: install the reported model, then repeat validation.

Stop or remove the local runtime

Stop local services:

mn runtime stop

For an installation made with the hosted installer, use its matching uninstall script:

curl -fsSL https://mirrorneuron.io/uninstall.sh | bash

Warning: ~/.mn can contain runtime configuration, run records, logs, and model metadata. Do not delete it until you have copied anything you need to retain.

Next steps

• Quickstart to validate and launch a model-backed local blueprint.
• Environment Variables to configure runtime state and listeners.
• Security Model before using real data or a shared network.