MirrorNeuron

MirrorNeuron is a local-first runtime for durable AI agent workflows. It is described as helping developers build super-agent workflows that can plan, research, run tools, and recover from failures on a local machine. The product is also presented as an open-source workflow runtime, released under the MIT License.

Its core workflow model centers on blueprints. Users can start with a blueprint instead of designing orchestration from scratch, run a blueprint with mock inputs, and then replace those inputs or adapters with their own code, data, or tools. The page also highlights reliability features for long-running work: retries, checkpoints, sleep, and resume are built in. MirrorNeuron is shown as supporting durable agent work that can keep making progress after interruptions, and the examples and use cases include finance, science research, and background agent workflows.

Deployment is local-first, with the same workflow shape able to move from a laptop to an edge node, private cluster, or cloud deployment. The page says workflows can run near files, feeds, tools, and systems that users already control, and it gives examples such as lab systems, market feeds, private files, device telemetry, customer records, and internal tools. Installation is shown through a CLI command, and the product includes a quickstart flow that installs the CLI, runs a blueprint, and then swaps in user code, data, or tools. A code example on the page uses Python and references the mn_sdk package, RunnerConfig.host_local(), and workflow recovery mode. MirrorNeuron is also described as a durable execution layer and workflow runtime rather than a hardware catalog.