Featured / CORE + ECOSYSTEM
The human side of agentic systems.
HRT is an open protocol and reference stack for integrating changing human capabilities into agentic systems.
Read the HRT field guide ↗
THE MISSING LAYER
The trace goes dark when a person enters the system.
Agents can trace models, tools, and outcomes. A person usually appears only as an approval, correction, or intervention.
HRT extends that operational record across the human–agent boundary—with time, uncertainty, and provenance—so existing systems can understand when human involvement changes the outcome.
How the protocol works ↗PARTNERS + EARLY ADOPTERS
Build the ecosystem from the role you already own.
HRT creates shared interfaces between verticals, processors, devices, and agent systems. Start with one bounded integration. Keep your implementation, data, and authority.
Explore partner onboarding ↗How Verticals Participate in HRT
Define how people, agents, tools, evidence, authority, and outcomes connect in your industry.
See the partner path ↗ Analytics providersHow Cognitive Processors and Analytics Providers Participate in HRT
Make a specialist capability discoverable, comparable, and composable while protecting its implementation.
See the partner path ↗ Device producersHow Sensor and Device Producers Participate in HRT
Make a sensor easier to integrate across processors and verticals without erasing its differentiation.
See the partner path ↗ Agent buildersHow Agent Systems Integrate HRT
Give agent systems a responsible interface to the human nodes their workflows already depend on.
See the partner path ↗ARTICLES + FIELD NOTES + GUIDES
Read the system from every angle.
10 pieces on protocol, trust, cognition, processor ecosystems, IP boundaries, and owner-operated testing.Human Runtime: A Stack for Human–Agent Integration
The open protocol, canonical runtime vocabulary, and shared infrastructure beneath HRT integrations and vertical products.
Read the field guide ↗ 0202 / HUMAN RELIABILITYCan Your AI Agent Trust the Human in the Loop?
When an agent asks a person, the trace goes dark. HRT makes the quality of that intervention legible.
Read the field note ↗ 0303 / AGENT INFRASTRUCTUREGive Your Agents a Human Trace
A protocol for connecting changing human state to agent networks—with confidence and provenance.
Read the field note ↗ 0404 / FAILURE ANALYSISBeyond the Artifact: When the Human in the Loop Fails
A person in the workflow is not automatically a safety system. Nine failures show where observability helps—and where it cannot.
Read the field note ↗ 0505 / DESIGNED JUDGMENTDesigned Judgment Needs a Runtime
Designed delegation defines who may decide. Human Runtime shows whether the decision point worked.
Read the field note ↗ 0606 / COGNITIVE OBSERVABILITYYou Do Not Need to Own the Mind to Observe the Runtime
What LLM observability reveals about tracing cognition when the underlying processor is not owned by the observer.
Read the research note ↗ 0707 / PROCESSOR ECOSYSTEMCompetition and Composability Between Processors While Protecting Their Implementations
Open processor slots, fair comparison, and protected implementations can reinforce one another.
Read the field note ↗ 0808 / IP BOUNDARIESWhen Open Interfaces Protect IP—and When They Expose It
What Datadog, OpenTelemetry, Khronos, prediction APIs, ONNX, and protected runtimes teach HRT about IP boundaries.
Read the research note ↗ 0909 / OWNER TESTHRT Processor IP Exposure and Extractability Self-Test
A private test that lets a processor owner measure leakage, extractability, and functional substitution before opening an interface.
Read the practical guide ↗ 1010 / METRIC SYSTEMHuman Runtime Cognitive Metrics
From cognitive vocabulary to an operational metric surface with observable primitives, provider estimates, and outcome evidence.
Read the technical reference ↗HRT TOPIC SORTER
Prime your LLM with HRT concepts—in minutes.
Explore one concept at a time. Grow a Topic Map grounded in your situation. Take it to your LLM.