Protocol v0.9 // Open Spec

Translate neural intent into structured API calls across any digital service. Three layers. One SDK. Zero vendor lock-in.

Read the DocsSee it in action
$ npm install @neuralintent/sdk
SmartHomeHealthKitMessagingCommerceNavigationMediaAccessibilityIoTRoboticsTelehealthAutomotiveEducationSmartHomeHealthKitMessagingCommerceNavigationMediaAccessibilityIoTRoboticsTelehealthAutomotiveEducation

Stripe for brain-computer interfaces

BCI hardware is advancing fast, but connecting neural signals to real-world services is still a nightmare of custom integrations. NeuralIntent provides the missing middleware layer: a universal translation protocol that turns thought into action.

BCI Hardware

EEG headsets, ECoG arrays, implanted electrodes. Any device that captures neural signals.

EEGECoGImplant

Digital Services

Smart home, messaging, commerce, healthcare, and any service with an API.

RESTGraphQLWebSocket

Three layers. One SDK.

Each layer handles one responsibility cleanly. Compose them together, or use them independently.

Layer 1

Intent Decoder

Translates raw neural signals from any BCI hardware into structured, classified intents. Supports EEG, ECoG, and implanted electrode arrays with sub-200ms latency.

import { IntentDecoder } from '@neuralintent/sdk';

const decoder = new IntentDecoder({
  hardware: 'eeg_generic',
  model: 'intent-v3',
  confidence_threshold: 0.85,
});

const intent = await decoder.classify(signal);
// => { type: 'navigation', action: 'scroll_down', confidence: 0.94 }
Layer 2

Service Router

Maps decoded intents to the correct API endpoint across any connected service. Plugin architecture supports custom service adapters with automatic discovery.

import { ServiceRouter } from '@neuralintent/sdk';

const router = new ServiceRouter({
  services: ['smart_home', 'messaging', 'commerce'],
  fallback: 'confirmation_prompt',
});

const route = router.match(intent);
// => { service: 'smart_home', endpoint: '/devices/control', method: 'POST' }
Layer 3

Confirmation Layer

Safety-critical verification step before any action executes. Supports visual, auditory, and haptic confirmation modes. Configurable per-action risk levels.

import { ConfirmationLayer } from '@neuralintent/sdk';

const confirm = new ConfirmationLayer({
  mode: 'visual',           // 'visual' | 'haptic' | 'auditory'
  risk_threshold: 'medium', // auto-confirm low-risk actions
  timeout_ms: 5000,
});

const approved = await confirm.request(route, payload);
// => { confirmed: true, method: 'neural_blink', latency_ms: 340 }

See intent-to-API in real time

Watch a neural intent get decoded, routed, and translated into a structured API call, step by step.

neuralintent-demo
🧠

Click "Run Demo" to see intent-to-API translation

<200ms
Intent decode latency
97.3%
Classification accuracy
40+
Service integrations
MIT
Protocol license

NeuralIntent reduced our BCI-to-app integration time from months to days. The intent decoder is remarkably accurate even with consumer-grade EEG hardware.

Dr. Elena Vasquez
CTO, Cortex Dynamics

The open protocol spec was what convinced us. We needed a standard our hardware partners could build against without vendor lock-in.

Marcus Lindgren
VP Engineering, Axon Interfaces

Our ALS patients can now control smart home devices with neural intent alone. The confirmation layer gives families peace of mind.

Dr. Priya Sharma
Director of Assistive Tech, Meridian Health

Start free, scale when ready

Open source core with commercial cloud and enterprise tiers.

Open Source

Free

MIT-licensed protocol spec and core SDK

  • NeuralIntent Protocol spec
  • Core SDK (TypeScript, Python)
  • 5 built-in service adapters
  • Community support
  • Local-only processing
Get Started

Developer

$0.002per intent

Cloud processing with managed infrastructure

  • Everything in Open Source
  • Cloud intent processing
  • 40+ service integrations
  • 99.9% uptime SLA
  • Priority support
  • Custom model fine-tuning
Start Building

Enterprise Medical

Custom

HIPAA-compliant, on-premise deployment

  • Everything in Developer
  • On-premise deployment
  • HIPAA/SOC2 compliance
  • Custom hardware support
  • Dedicated success engineer
  • Clinical validation tools
Contact Sales

Frequently asked questions

What BCI hardware does NeuralIntent support?
NeuralIntent is hardware-agnostic. Our SDK includes adapters for major EEG headset manufacturers, ECoG systems, and implanted electrode arrays. The open protocol spec means any hardware vendor can build a compatible adapter. We are actively partnering with FDA-approved implant companies.
How does the Intent Decoder handle ambiguous signals?
The decoder uses a confidence threshold system. Signals below the configurable threshold are routed to the Confirmation Layer for explicit user verification rather than guessing. You can tune sensitivity per-use-case, from high-confidence-only for medical scenarios to more permissive for casual smart home control.
Is this safe for medical use?
Our Enterprise Medical tier is designed for clinical environments with HIPAA compliance, audit logging, and configurable safety overrides. The Confirmation Layer ensures no action executes without appropriate verification. We work with clinical partners to validate safety protocols.
What is the NeuralIntent Protocol?
The NeuralIntent Protocol is an MIT-licensed open specification that defines a standard format for neural intent data, service routing rules, and confirmation flows. It enables interoperability between different BCI hardware, middleware, and service providers without vendor lock-in.
How do I build a custom service adapter?
Service adapters implement a simple interface: they receive a classified intent and return an API payload. Our SDK provides a base adapter class with automatic discovery, health checks, and retry logic. See the docs for a step-by-step guide that takes about 30 minutes.
What are the latency characteristics?
End-to-end from neural signal to API call execution is typically under 500ms. The Intent Decoder processes in under 200ms, Service Router matching is under 5ms, and the remainder depends on the target service response time. Cloud processing adds approximately 50ms over local-only.

Build the future of human-computer interaction

Join the developers, researchers, and healthcare teams building on the NeuralIntent Protocol.

Start BuildingView Protocol Spec