Pinned package consumption
Downstream products can consume world-runtime as a supported package and start it through the released CLI or module serve entrypoints.
v1.1.0 GA
a governed world-state runtime for consequential enterprise decisions.
world-runtime gives humans, software, and AI agents an explicit model of operating reality: durable state, append-only events, simulation before commitment, policy-governed action, and audit-ready replay.
Enterprise systems already store the truth. Models already generate intent.
The missing layer is the runtime between AI intent and real-world consequence. world-runtime coordinates state, events, policies, approvals, simulations, and operator control across enterprise decision systems.
Governed state over time
Track entities, relationships, events, proposals, and decisions instead of treating every AI interaction as an isolated request.
Policy, approvals, and provenance
Keep human review, rule evaluation, tool boundaries, and decision evidence inside the runtime itself by design.
Reusable substrate
Build multiple decision systems on one shared runtime instead of rebuilding the same governance layer for every workflow.
Domain Agents & Apps
world-runtime
Systems of Record
Humans, Policy & Observability
What is new in v1.1.0
The runtime is now easier to consume, supervise, route through, inspect, and validate.
The v1.1.0 release keeps the v1 App Server, Public API, SDK, persistence, and extension-contract posture intact while adding downstream-consumable managed runtime capabilities.
Managed runtime services
Local runtime-adjacent services can be described, supervised, health-checked, restarted, and inspected through validated service manifests.
Provider and task routing
Teams can map structured extraction, assistant, and related workloads to provider inventory with deterministic routing and bounded fallback.
Runtime admin surfaces
Inventory, provider inspection, task resolution, and bounded reconcile are exposed through supported App Server, Public API, and SDK paths.
Structured extraction reference stack
A local AI reference path shows schema-shaped extraction workloads validated through the released runtime management surfaces.
Why
The missing layer is not more intelligence.
It is governed consequence.
AI can propose action faster than enterprises can govern consequence. Teams need explicit world state, policy, approvals, simulation, and replay between intent and commitment.
Shared truth deserves shared runtime logic.
Build
Public proof paths show the runtime under consequential domain pressure.
Learn how to turn the runtime into concrete domains: supply networks, lab science, air traffic, power grids, city operations, markets, autonomous systems, and multi-agent coordination.
Simulation, replay, branching, and policy stay explicit.
How
Internal teams can keep their stack while giving decisions a governed world underneath.
Works with your stack: keep your models, tools, services, and cloud choices, then use world-runtime to structure state, proposals, simulation, policy, approvals, and audit-ready replay.
Governed build loops without a prescribed stack.
Docs
Canonical entrypoints keep evaluation connected to implementation evidence.
Docs focuses the repo into usable paths: evaluator orientation, v1.1.0 release notes, consumer guidance, developer quickstarts, public APIs, operator playbooks, and extension contracts.
The repository stays the source of truth.
Repository
Explore world-runtime
Further reading
The Orchestrated Enterprise
Robb Bush lays out the business-strategy logic behind world-runtime: why coordination, trust, governance, and orchestration-native operating models matter more as intelligence becomes abundant.
Explore the book
Business strategy behind world-runtime