Tyto Sensory Labs

Multimodal Communications Protocol For Humanity

Preserve meaning. Adapt the carrier.

MCP4H is Tyto's framework for a simple problem: useful information often exists in a form the receiver cannot use directly.

Instead of forcing every source, device or person to speak the same language, MCP4H aims to preserve what the information means, then adapt how it is expressed.

ObserveInterpretChoose a projectionDeliverPerceiveRespond

In plain English

The source should not dictate how the message must be received.

A machine may expose telemetry. A person may need touch. Another receiver may need sound, force, light, text, spatial cues or a machine-readable event.

MCP4H sits between those ends. It tries to keep the useful meaning intact while allowing the carrier to change.

Simple example

A simulator reports tyre and chassis state as numbers. MCP4SH interprets those numbers as useful driving events. A tactile system can then express those events through a shaker. A future wheelbase renderer could express the same useful state differently without changing what the event means.

What stays separate

Meaning is easier to reuse when we stop mixing it with the plumbing.

Observation

What the source actually measured or reported.

Interpretation

What that observation appears to mean in context.

Policy

Optional decisions about priority, suppression, escalation or timing.

Projection

How the meaning should be expressed for this receiver.

Transport

How the information gets from one part of the system to another.

Rendering

The final physical or digital form: touch, sound, force, light, text or something else.

Patterns over time

A number is not always the message.

Useful state can live in how something changes: rate, rhythm, oscillation, frequency, phase, modulation, recurrence, direction or relationships between signals.

MCP4H therefore needs to preserve meaningful temporal, spatial, relational and spectral structure when a task depends on it. Reducing everything to a static number or a sentence can throw useful information away.

The receiver matters

The best carrier depends on what is available at the other end.

A receiving system may have different senses, hardware, actuator positions, usable frequency ranges, latency limits, accessibility needs, preferences or context.

  • Same meaning, different senses: an alert that would normally be heard could be translated into touch or light.
  • Same meaning, different hardware: a vehicle event could be expressed through a shaker, pedal motor or wheelbase detail layer.
  • Same meaning, different context: a cue that is useful while driving may need a different presentation when reviewing data later.
Receiver capability is not an afterthought. It is part of deciding how the message should be expressed.

AI and interpretation

AI can help understand complexity; it is not required for the protocol to work.

Some signals can be interpreted deterministically. Others may benefit from classifiers, learned models or AI. MCP4H should support both.

Where interpretation is uncertain, the framework should keep useful provenance and confidence so a receiver can tell where the conclusion came from and how strongly it is supported.

Where it is being tested

Generality has to be earned.

  • Simulation: MCP4SH is the first commercial reference implementation. Racing is the established proving ground; flight broadens the source domain.
  • Remote operation: a future direction for restoring machine and environmental cues that disappear when the operator is somewhere else.
  • Accessibility: a future reference domain for translating information away from a channel that is unavailable or overloaded.
  • People + AI: a way to make useful machine interpretation perceivable without turning every conclusion into more text.
  • Other living systems: a longer-horizon research direction where we should not assume our language or senses are the only useful representation.

Status

An active framework, not a finished universal standard.

MCP4H is being developed in public. The next work focuses on a neutral semantic envelope, receiver capability profiles, provenance, temporal and spectral structure, domain profiles and reference implementations that prove the same base architecture can work across genuinely different problems.

If the framework only works naturally for the problem it started with, it is not general enough yet.