Human Runtime, or HRT, is an open protocol and infrastructure layer for connecting changing human capabilities to agentic systems. It gives human-performance observations, derived states, task events, agent actions, interventions, and outcomes a shared operational structure.

The central ecosystem proposition is:

HRT creates competition and composability between capabilities while protecting the implementations that make each participant valuable.

HRT is not asking one company to build or disclose the whole stack. It creates defined roles through which different companies can contribute and benefit.

Why this matters now

Software is moving from applications that wait for instructions toward systems that plan, call tools, exchange tasks, monitor outcomes, and ask people for help. The exact pace will differ by industry. Full autonomy is not required for the shift to matter.

As systems become more agentic, the human participant becomes a runtime dependency:

  • an expert reviews an uncertain result;
  • an operator receives a handover;
  • a supervisor authorizes an action;
  • a worker supplies context that the agents lack;
  • a trainer evaluates an attempt;
  • a customer or citizen is asked to correct the system;
  • a person must notice when automation has failed.

Today, the trace usually becomes incomplete at that point. The system records that a person approved, rejected, corrected, or intervened. It rarely records the observable conditions under which the contribution was produced, whether the information reached the person in time, or whether the intervention improved the outcome.

HRT supplies that missing coordination and evaluation layer.

One architecture, several roles

Verticals

A vertical defines what matters in its environment: tasks, cues, goals, risks, authority, outcomes, and intervention rules. HRT does not reduce aviation, industrial work, healthcare, mobility, training, professional review, or other fields to a generic cognitive score.

Cognitive processors and analytics providers

Processors turn authorized signals and context into documented measurements, states, or forecasts. They can participate through a common interface, remain proprietary, and be compared against independent outcomes.

Sensor and device producers

Sensors create observations and quality metadata. They can become easier for processors, applications, and new verticals to use without giving HRT their firmware or signal-processing implementation.

Agent systems

Agent platforms can discover HRT-aware capabilities, request authorized human context, coordinate handovers, select processors, and connect human interventions to the machine trace.

Observability and infrastructure providers

Infrastructure companies can transport, store, query, visualize, redact, route, and evaluate HRT data while respecting its additional consent and governance requirements.

Participation is not one commitment

An organization can participate at several depths.

  1. Adopter — uses an existing HRT profile or adapter.
  2. Compatible provider — exposes a tested HRT interface around a capability.
  3. Reference provider — maintains a high-quality implementation and example.
  4. Vertical partner — defines a domain profile, workflow, and evaluation method.
  5. Research or evaluation partner — contributes approved data, test methods, or outcome evidence.
  6. Steward — helps govern shared semantics, conformance, or safety requirements.
  7. Infrastructure supporter — supplies engineering, hosting, funding, or distribution.

An adopter gains compatibility without becoming responsible for the standard. A vertical partner has more influence because it contributes domain structure and evidence.

What remains open and what remains owned

The open HRT layer can define:

  • Human Nodes, Runtime States, Runtime Spans, Runtime Events, and Human Traces;
  • time, provenance, uncertainty, validity, and authorization fields;
  • processor and device profiles;
  • vertical task and outcome profiles;
  • evaluation traces and conformance tests;
  • A2A, OpenTelemetry, and other ecosystem bindings.

Participants can retain:

  • algorithms, models, weights, thresholds, and feature engineering;
  • firmware and device differentiation;
  • private datasets and customer information;
  • applications, workflows, dashboards, and commercial services;
  • provider-specific extensions;
  • customer relationships and vertical expertise.

The open layer makes capabilities legible and connectable. It does not make their implementations common property.

Why join before the architecture settles

Early participants can help define the interfaces through which their category will later be discovered, evaluated, and integrated. That provides several advantages:

  • less repeated integration work;
  • a visible place in an emerging market map;
  • earlier access to new vertical demand;
  • influence over scientifically and commercially realistic requirements;
  • a reusable evidence and evaluation surface;
  • a route into agent systems without building an entire agent platform;
  • clearer differentiation based on demonstrated capability rather than proprietary integration friction.

The objective is not to nominate one permanent winner. It is to make serious providers easier to find and their differences easier to evaluate.

A practical starting point

Every partnership can begin with one bounded profile and one trace:

  1. Choose a real or synthetic workflow.
  2. Identify the participating device, processor, vertical, agent, and outcome roles.
  3. Map only the minimum authorized signals and outputs.
  4. Create an HRT trace connecting the episode to its result.
  5. Run conformance, usefulness, and exposure tests.
  6. Decide whether the relationship remains private, becomes a reference integration, or expands into a pilot.

HRT can build the adapter, trace representation, and agent-facing interface. The partner supplies only the domain or capability knowledge required to make the profile valid.

The shared case for participation

Your company does not need to become a complete human–agent platform. It can expose the capability it already owns through a controlled HRT interface. HRT makes that capability discoverable, composable, attributable, and connected to tasks and outcomes. Other providers can participate too, which creates genuine comparison and a larger market. Your implementation, data, extensions, and commercial services remain yours. You choose the depth of each relationship. HRT supplies the shared architecture through which the ecosystem can grow.

Role-specific adoption paths