Why MirrorNeuron
Decide whether MirrorNeuron fits a local or clustered AI workflow.
Why MirrorNeuron
Use this page if you are evaluating MirrorNeuron.
What MirrorNeuron is for
MirrorNeuron is a fit when you want:
- Easy: Run long or complex AI jobs without worrying about which model, machine, or runtime should handle each step. Reuse proven workflows instead of rebuilding them from scratch.
- Control: See what the AI is working on, what actions it has taken, and what evidence supports its decisions. Pause, resume, cancel, or recover jobs whenever needed.
- Ownership: Keep your workflows, runtime state, and outputs under your control. Run work locally, inside private sandboxes, or across your own runtime nodes, while connecting external services only when you choose.
What you should evaluate first
Before adopting MirrorNeuron, consider four questions:
| Question | Why it matters |
|---|---|
| Why do you want to run AI locally or at the edge? | Local execution gives you more control over sensitive data, latency, cost, and system availability. It is most valuable when your workflows need to stay close to private files, factory systems, engineering tools, or other resources that should not depend entirely on the cloud. |
| Do you have the hardware required for your workload? | Some MirrorNeuron blueprints can run on a MacBook, but production workloads may require more memory, faster GPUs, or dedicated machines such as a Mac Studio, high-end workstation, or NVIDIA DGX Spark. Hardware needs depend on the models, tools, and workload size used by each blueprint. |
| Do you need a deep AI agent rather than a general AI assistant? | Tools such as coding agents and AI coworkers are useful for interactive tasks. A deep agent is designed for complex jobs that may run for hours or days, combine multiple LLM calls with conventional software, simulations, databases, and specialized tools, and recover from interruptions while working toward a long-horizon goal. |
| Why should AI move closer to your data—not the other way around? | Video, sensor, and physical-AI workloads generate large, continuous data streams and often require immediate responses. Sending everything to the cloud can add latency, bandwidth costs, and reliability risks. For these workloads, edge AI is not only about privacy; it may be the only practical way to operate in real time. |
What MirrorNeuron does not guarantee by itself
MirrorNeuron gives you more control over where AI runs and how data is handled, but it does not make every workflow private, secure, or correct automatically.
-
Local AI does not always mean a fully offline workflow. Your AI models may run entirely on your own machine, while skills such as web browsing, email, APIs, or external connectors may still access the internet. For stricter isolation, you can run MirrorNeuron in a fully air-gapped environment and verify that the workflow continues operating without any network connection.
-
Open source gives you verifiability, not a security guarantee. MirrorNeuron does not come with a managed-service security agreement or contractual guarantee. Instead, you can inspect the code, review blueprint permissions, control network access, and run the system entirely on infrastructure you own. For a fully local workflow, you can disconnect the machine from the internet and verify that the job continues running.
-
Sandboxing reduces risk but does not remove it. Our sandbox, NVIDIA OpenShell, provides a stronger execution boundary than running directly on the host machine, but operators must still review file access, uploaded content, environment variables, network policies, and available services.
-
Successful execution does not guarantee a correct decision. A completed blueprint run means the workflow executed successfully; it does not make the result authoritative. Our workflows in blueprints are designed to produce evidence and review material, but not replace qualified human judgment.
Next steps
- Installation for a local runtime.
- Quickstart for a model-backed local blueprint preflight and run.
- Core Concepts for the workflow, blueprint, job, and runtime-node vocabulary.
- Security Model before using real data or a shared network.