Skip to content

Feature Request: Robot Task Protocol (RTP) for Task Commissioning and Payment #586

@plagtech

Description

@plagtech

Feature Request: Robot Task Protocol (RTP) for Task Commissioning and Payment

Problem

Isaac GR00T enables developers to train and deploy humanoid robot foundation models. But once a robot is deployed and operational, there's no standardized protocol for external AI agents to discover, commission, and pay robots for physical tasks in real-time.

Today, if an AI agent running in OpenShell or NemoClaw needs a robot to pick up a package, take a sensor reading, or run an assembly task — there's no HTTP-native way to:

  1. Discover which robots are available and what they can do
  2. Negotiate pricing for a specific task
  3. Pay the robot operator with a verifiable micropayment
  4. Receive proof of task completion

This gap becomes critical as GR00T-powered robots move from labs into production fleets where they need to interoperate with autonomous AI agent systems.

Proposed Solution

The Robot Task Protocol (RTP) is an open standard for AI agents to commission robots via HTTP and pay with USDC micropayments using the x402 protocol.

How RTP Works

AI Agent                    RTP Gateway                  Robot/Device
   |                            |                            |
   |-- POST /rtp/discover ----->|                            |
   |<-- Available devices ------|                            |
   |                            |                            |
   |-- POST /rtp/hire --------->|                            |
   |   (includes x402 payment)  |                            |
   |                            |-- Task assignment -------->|
   |                            |                            |
   |                            |<-- Task result ------------|
   |<-- Result + proof ---------|                            |

RTP Endpoints

Endpoint Description
GET /v1/rtp/discover List available robots/devices with capabilities and pricing
POST /v1/rtp/hire Commission a task with x402 USDC micropayment
GET /v1/rtp/status Check task execution status
POST /v1/rtp/cancel Cancel a pending task

Resources

  • RTP Spec: https://github.com/plagtech/rtp-spec
  • RTP SDK: https://github.com/plagtech/rtp-sdk
  • RTP Python SDK: https://github.com/plagtech/rtp-python-sdk
  • Live gateway (8 RTP endpoints): https://gateway.spraay.app
  • x402 Protocol: https://www.x402.org
  • Pi 5 hardware demo: In progress for Base Batches 2026 Robotics Track

Alternatives Considered

  • Custom REST APIs per robot vendor — no standardization, every integration is bespoke
  • ROS 2 service calls — not HTTP-native, can't integrate with web-based AI agents
  • DePIN protocols (peaq, Auki) — complementary infrastructure layers, not task-commissioning protocols

Who Built This

RTP is developed by @plagtech, the same team behind the Spraay x402 payment gateway. We've already submitted integrations to:

# Feature Request: Robot Task Protocol (RTP) for Task Commissioning and Payment

Problem

Isaac GR00T enables developers to train and deploy humanoid robot foundation models. But once a robot is deployed and operational, there's no standardized protocol for external AI agents to discover, commission, and pay robots for physical tasks in real-time.

Today, if an AI agent running in OpenShell or NemoClaw needs a robot to pick up a package, take a sensor reading, or run an assembly task — there's no HTTP-native way to:

  1. Discover which robots are available and what they can do
  2. Negotiate pricing for a specific task
  3. Pay the robot operator with a verifiable micropayment
  4. Receive proof of task completion

This gap becomes critical as GR00T-powered robots move from labs into production fleets where they need to interoperate with autonomous AI agent systems.

Proposed Solution

The Robot Task Protocol (RTP) is an open standard for AI agents to commission robots via HTTP and pay with USDC micropayments using the x402 protocol.

How RTP Works

AI Agent                    RTP Gateway                  Robot/Device
   |                            |                            |
   |-- POST /rtp/discover ----->|                            |
   |<-- Available devices ------|                            |
   |                            |                            |
   |-- POST /rtp/hire --------->|                            |
   |   (includes x402 payment)  |                            |
   |                            |-- Task assignment -------->|
   |                            |                            |
   |                            |<-- Task result ------------|
   |<-- Result + proof ---------|                            |

RTP Endpoints

Endpoint Description
GET /v1/rtp/discover List available robots/devices with capabilities and pricing
POST /v1/rtp/hire Commission a task with x402 USDC micropayment
GET /v1/rtp/status Check task execution status
POST /v1/rtp/cancel Cancel a pending task

Device Categories

Category Examples
robotics Humanoid robots (GR00T-powered), robotic arms, manipulators
sensing Environmental sensors, cameras, LiDAR arrays
delivery Drones, autonomous vehicles, couriers
manufacturing 3D printers, CNC machines, assembly stations
compute Edge GPU nodes, inference endpoints

Integration with Isaac GR00T

A GR00T-powered robot could expose its capabilities via RTP:

# Robot registers its capabilities on the RTP network
rtp_registration = {
    "device_id": "gr00t-neo-001",
    "model": "Isaac GR00T N1.6",
    "embodiment": "1X NEO Gamma",
    "capabilities": [
        {"task": "pick_and_place", "price_usdc": "0.05"},
        {"task": "object_sorting", "price_usdc": "0.10"},
        {"task": "shelf_stocking", "price_usdc": "0.25"},
    ],
    "location": "warehouse-a",
    "status": "available"
}

An AI agent in OpenShell can then discover and hire this robot:

# Agent discovers available GR00T robots
devices = await rtp_discover(category="robotics")

# Agent hires a robot for a specific task
result = await rtp_hire(
    device_id="gr00t-neo-001",
    task="pick_and_place",
    parameters={"object": "red_box", "destination": "shelf_b"},
    max_budget="0.05"  # USDC
)

Why x402 for Robot Payments

  • Micropayments: Tasks can cost fractions of a cent — traditional payment rails can't handle this
  • No accounts needed: Robots and agents pay per task, no subscription or invoicing
  • On-chain verification: Both parties have cryptographic proof of payment and task completion
  • Multi-chain: Works across Base, Ethereum, Solana, and 10 other chains

How This Connects to the Isaac Ecosystem

NVIDIA Component RTP Role
GR00T N1.6 Robot foundation model — RTP exposes its capabilities as hireable services
Isaac Sim Simulation — test RTP task commissioning in simulation before real deployment
Isaac Lab-Arena Benchmarking — evaluate robot performance on RTP-commissioned tasks
OSMO Orchestration — coordinate RTP-enabled robot fleets across environments
Jetson Thor Edge compute — run RTP agent on the robot's onboard computer

Resources

Alternatives Considered

  • Custom REST APIs per robot vendor — no standardization, every integration is bespoke
  • ROS 2 service calls — not HTTP-native, can't integrate with web-based AI agents
  • DePIN protocols (peaq, Auki) — complementary infrastructure layers, not task-commissioning protocols

Who Built This

RTP is developed by [@plagtech](https://github.com/plagtech), the same team behind the Spraay x402 payment gateway. We've already submitted integrations to:

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions