MirrorNeuron
an open-source runtime for reliable on-device AI agents
Last verified:
What is MirrorNeuron?
MirrorNeuron is an open-source runtime for durable local AI workflows and agent orchestration. It allows users to create workflows from reusable blueprints and run them on a laptop, cluster, edge node, or cloud without changing the workflow concept. The platform handles scheduling, state persistence, retries, backpressure, and cluster failover automatically, providing a simple path from first working blueprint to reliable background workflows.
Key features include a CLI interface with commands like validate, run, status, monitor, and blueprint run, plus a Python SDK for authoring workflows with decorators and an HTTP API for programmatic job control. MirrorNeuron separates the control plane from the execution plane, using the BEAM runtime to orchestrate agents and track job state in Redis while workers run in sandboxed or local environments. The tool supports multi-agent workflow graphs with routers, executors, and aggregators.
MirrorNeuron is designed for developers and teams building AI agents and multi-agent systems who want to avoid orchestration complexity like Airflow DAG sprawl or Temporal cluster complexity. It's suitable for marketing workflows like personal email outreach campaigns, technical tasks like prime number sweeps, and any long-running agent work that needs reliability with built-in retries, checkpoints, and resume capabilities.
The platform is open source on GitHub under an MIT license, making it accessible for teams who want to inspect, adapt, and share workflows clearly. Users can start by installing the CLI with a single curl command, running an example blueprint, and gradually customizing workflows for their specific needs.
MirrorNeuron pricing
Pricing model: Freemium
MirrorNeuron is completely free and open source under an MIT license on GitHub. There are no paid tiers, subscription plans, or enterprise features. Users get full access to the CLI, Python SDK, HTTP API, blueprints, and cluster capabilities without any cost.
MirrorNeuron pros
- Open source with MIT license on GitHub
- Run workflows on laptop, cluster, edge, or cloud
- Built-in retries, checkpoints, and resume for long-running work
- Start in minutes with reusable blueprints
- No Airflow DAG sprawl or Temporal cluster complexity
- Automatic scheduling, state persistence, and cluster failover
- CLI, Python SDK, and HTTP API for workflow authoring
- BEAM runtime orchestrates agents automatically
- State tracked in Redis for persistence
- Workflows defined as graphs of routers, executors, aggregators
- Share and inspect workflows with team members
- Built-in backpressure handling
- Validate bundles before running with mn validate
- Live event monitoring with mn monitor command
- Job pause and resume capabilities
- Cluster setup across multiple machines with failover
- Sandboxed or local worker environments via OpenShell or HostLocal
MirrorNeuron cons
- Requires Redis installation and configuration
- gRPC runtime must be reachable for CLI commands
- Primarily Linux/WSL focused installation
- Cluster setup requires helper scripts for controlled startup
- mn clear permanently deletes production data from Redis
- Rich terminal formatting can corrupt script output in CI
- Default timeout of 10 seconds may be too short
- Documentation scattered across multiple pages
Frequently asked questions about MirrorNeuron
What is MirrorNeuron?
MirrorNeuron is an open-source runtime for durable local AI workflows and agent orchestration. It runs durable multi-agent workflows on one machine or across a cluster, handling scheduling, state persistence, retries, backpressure, and cluster failover automatically.
How do I install MirrorNeuron?
Install by running the command: curl -fsSL https://mirrorneuron.io/install.sh | bash. This installs the CLI and starts local services. The CLI uses mn-python-sdk to communicate with the MirrorNeuron core over gRPC.
What are blueprints in MirrorNeuron?
Blueprints are reusable, working AI workflows that users can run immediately with one command. Examples include business_email_campaign_deamon for marketing outreach and general_test_message_flow for testing. Users start from a blueprint, run one command, and customize later.
How do I run a workflow?
First validate with mn validate <bundle_path> to check manifest structure, then run with mn run <bundle_path> to submit the job. MirrorNeuron schedules agents, routes messages, and persists state in Redis. Use mn status and mn monitor to watch progress.
Can MirrorNeuron run across multiple machines?
Yes, MirrorNeuron runs across a cluster with automatic job failover. Use mn join <ip> to connect to a node and mn leave <node_name> to disconnect. Cluster helper scripts handle node naming, port configuration, and Redis HA setup in the correct order.
What programming languages does MirrorNeuron support?
MirrorNeuron primarily supports Python through its Python SDK which uses decorators for authoring workflows. Workers can also be Python scripts, shell commands, or any executable running in sandboxed or local environments.
How does MirrorNeuron handle failures?
MirrorNeuron has built-in retries, checkpoints, sleep, and resume capabilities for long-running agent work. It handles cluster failover automatically, tracks job state in Redis, and provides dead-letter events for messages that could not be routed or processed.
What is the BEAM runtime?
The BEAM runtime is MirrorNeuron's control plane that orchestrates agents, tracks job state in Redis, and routes messages between nodes. It separates the control plane from the execution plane where workers actually run.
How do I monitor a running job?
Use mn monitor <job_id> to stream live events for a running job in real-time. It keeps streaming until the job reaches a terminal state (completed, failed, or cancelled) or you interrupt it. Also use mn status for current job record and mn result for artifacts.
Is MirrorNeuron free to use?
Yes, MirrorNeuron is completely open source on GitHub under an MIT license. There are no paid plans, subscriptions, or enterprise tiers. All features including CLI, Python SDK, HTTP API, blueprints, and cluster capabilities are free.