The operational command center for governed loops.
Millrace OS is WiP, estimated release by July 1, 2026. It is the planned dashboard for seeing which Millrace daemons are active, inactive, ready, blocked, recovering, or complete across the workflows you operate.
Operate daemons without turning the dashboard into the runtime.
Millrace OS is intended to make daemon operation legible: what is queued, what is running, why a workflow is blocked, what evidence exists, what is complete, and whether each daemon is ready to accept more work.
Queues
Watch intake, runnable work, blocked records, recovery routing, and pending completion checks across governed loops.
Runs
Follow stage assignments, terminal markers, evidence artifacts, and deterministic transitions as runs move through the compiled graph.
Readiness
See daemon health, configuration status, available runners, stalled work, and operator actions needed before more dispatch.
APIs and connectors should make Millrace manageable from wherever work starts.
Expose daemon status, queues, runs, blocked states, evidence, completion, and readiness through a control API rather than scraping terminal output.
Let external tools and agents enqueue work, inspect progress, and respond to blocks through governed operations.
Keep the control surface stable enough that teams can build their own views without forking the runtime or depending on one bundled interface.
Local operation needs local-model discipline.
Millforge is the offline/local-model path: stage-specific guardrail configurations around a custom harness runner, intended to improve smaller-model reliability by deterministically blocking invalid tool use.