Run dependable AI workflows on your machines

Bring your work. MirrorNeuron handles execution.

View on GitHub
MIT licensedmacOS, Linux & WSL2Docker required
No cloud required
How it works

Bring your work. Get a runtime.

Turn working code into a dependable workflow — without building the execution stack yourself.

Your work
Code · Agents · Models · Tools
MirrorNeuron
Sandbox
Model setup
Context
Checkpoints
Recovery
Human gates
Resources
Observability
Your compute
macOS · Linux · WSL2
one machine or cluster
What this enables
01 / Dependability
Reliable workflow execution.
Durable state, checkpoints, and recovery keep long-running work moving through failures.
02 / Easy deployment
Ready in one command.
One command sets up the runtime, models, and dependencies you need.
03 / Control
Cloud optional.
Keep execution on your machines, with cloud services optional and workflows portable.
Where it fits

Use the smallest runtime that solves the problem.

Start simple. Add infrastructure only when the work demands it.

Python script
The job is short and restarting is cheap
Agent framework
You're designing agents, prompts, tools, and control flow
Temporal / Airflow
You need general-purpose workflow orchestration
MirrorNeuron
Long-running AI work needs dependable execution on infrastructure you control

Work that outlives a chat.

Long-running AI work that needs to survive failures, adapt, or stay local.

Quickstart

Run your first workflow.

Start from a working blueprint, inspect the execution, then replace the example logic with your own code.

bash — mirrorneuron quickstartready
>_1. Install MirrorNeuron
$ curl -fsSL https://mirrorneuron.io/install.sh | bash
>_2. Run a resilient blueprint
$ mn blueprint run vc_assistant