Governance for robots that think.
A control plane that decides what a robot is allowed to do — and enforces it before the action executes, not a dashboard of what already happened. Safety, compliance, and intent gated at runtime across industrial fleets and domestic robots, through open protocols.
Member of —
Robots are shipping faster than the rules that govern them
robots deployed globally
Industrial robots in active operation worldwide (IFR World Robotics 2024)
runtime governance standards
ISO 10218, IEC 61508, ISO 13482 protect humans from robots. None cover AI decision governance.
open standard — and we built it
AIRGP is the first unified runtime AI governance protocol
ThinkNEO built the missing layer.
We don't replace anything. We add what's missing.
ThinkNEO is an optional governance layer that plugs into your existing robot stack. Your vendor keeps full ownership of hardware and software. We only add the decision governance that no vendor provides today.
Manufacturer owns the hardware
Motors, sensors, chassis, actuators — built and warranted by the robot maker.
Manufacturer owns the system
Firmware, OS, drivers, ROS stack — maintained and updated by the OEM.
ThinkNEO owns the decision layer
Runtime governance, policy enforcement, intent verification, audit trails — every AI decision governed in real time.
You don't control the robot.
You govern what it decides to do.
Why multi-robot governance is possible
Every robot runs AI models to make decisions
Same layer, any vendor
AI inference is distributed — cloud, edge, or hybrid
Accessible via standard APIs
The decision layer can be unified across all robots
One control plane, every fleet
ThinkNEO is an optional layer — it doesn't touch motors, firmware, or hardware. It only influences the AI decisions that drive robot behavior. Add it to any robot, from any manufacturer. Remove it anytime. Zero lock-in.
Your vendor builds the robot. ThinkNEO is the optional safety net for its decisions.
Mission Control for your robot fleet
Open protocols, not a black box.
ThinkNEO authored AIRGP and A2ASTC as open specifications. Implement them, audit them, extend them — your fleet is never locked in.
AIRGP
Artificial Intelligence Runtime Governance Protocol
An open standard for real-time policy enforcement on AI agents. Define rules in human-readable YAML, deploy to edge, and enforce in microseconds.
A2ASTC
Agent to Agent Software Team Compliance
A cryptographic protocol for multi-agent coordination. Robots verify each other's intent and compliance status before collaborative actions.
ThinkShield
Open Defense Layer for AI Governance Platforms
Production defense engine with 5 rule packs, detection scoring, and ASGI middleware. Protects every request across the ThinkNEO platform. 179 tests, p99 0.825ms.
Stewarded by
ThinkNEO - Hong Kong
From the factory floor to the living room
Manufacturing
Enforce safety envelopes on cobots collaborating with humans on the line.
Logistics
Govern AMR fleets across multi-vendor warehouses with a single policy plane.
Energy & Utilities
Audit autonomous inspection drones against regulatory frameworks.
Scenarios shown reflect platform design targets; deployment metrics published as customers go on record.
Connect any robot in minutes, not months
Lightweight SDKs, robotics-specific APIs, and native adapters that plug into your existing stack. Nothing to replace — ThinkNEO adds a governance layer on top of what you already run.
from thinkneo import GovernanceClient
# Connect to ThinkNEO control plane
client = GovernanceClient(
api_key="tn_live_...",
fleet_id="fleet_warehouse_01"
)
# Before any robot action — verify intent
verdict = client.verify_intent(
robot_id="AMR-WH01-0042", # warehouse AMR (any vendor)
action="navigate",
parameters={"destination": "zone_b", "speed": 0.8},
context={"humans_nearby": 2}
)
if verdict.allowed:
robot.execute(verdict.action) # Hardware executes
else:
robot.halt() # Policy blocked
log.warning(verdict.reason) # "speed > 0.5 when humans_nearby > 0"Official SDKs
Python SDK — request early accessTypeScript SDK — request early accessROS 2 node thinkneo_governance — request early accessRobotics-Specific APIs
Intent Verification API
<0.4msSubmit robot actions for real-time policy evaluation before execution.
POST /v1/intent/verifyFleet Telemetry API
real-timeStream sensor data, battery, position, and operational metrics.
WS /v1/telemetry/streamAudit Chain API
<10msQuery and export immutable decision logs for compliance and insurance.
GET /v1/audit/chainPolicy Manifest API
<50msDeploy, version, and sign governance policies across your fleet.
PUT /v1/policies/deployEmergency Control API
<1msRemote kill-switch and emergency override for any governed robot.
POST /v1/emergency/stopMulti-Agent Coordination
<2msA2ASTC-compliant inter-robot communication and task verification.
POST /v1/a2astc/coordinateHardware & Framework Adapters
ROS 2 Adapter
BetaNative ROS 2 node — subscribes to /cmd_vel, /joint_states, publishes governance verdicts
NVIDIA Isaac Adapter
BetaIsaac Sim / Isaac ROS integration via OmniGraph extension. Reference integration available to design partners.
Custom Humanoid Stacks
GARK3588-based and other custom humanoid platforms integrate via the native C++/Python SDK or the gRPC adapter — no ROS dependency required.
gRPC Adapter
BetaLanguage-agnostic, high-performance binary protocol for custom stacks
REST Adapter
BetaHTTP/JSON for lightweight integrations, webhooks, and legacy systems
MQTT Adapter
BetaIoT-native protocol for edge devices with constrained bandwidth
WebSocket Adapter
BetaBidirectional real-time streaming for dashboards and monitoring
How to add ThinkNEO to any robot — no changes to existing systems
Add the adapter
Drop-in ROS 2 node, Isaac extension, or lightweight SDK — no changes to your existing code
Register fleet
Each robot authenticates and receives a policy manifest — your vendor stack stays untouched
Governance activates
AI decisions pass through AIRGP verification (<0.4ms) — optional per robot, per action, per policy
Full auditability
Immutable decision logs — remove ThinkNEO anytime and your robot keeps working as before
One control plane. Every robot.
Operator Console
Central dashboard for fleet management and monitoring
AIRGP Engine
Intent verification in microseconds
A2ASTC Bus
Multi-agent coordination layer
Robot Fleet
Built for regulators, insurers, and engineers.
Audit-ready
Every robot decision is logged, hashed, and chained. Export for any audit framework.
Standards-aligned
Designed against ISO 10218, IEC 61508, EU AI Act high-risk Annex III, and emerging ISO/IEC 42001 AIMS.
Open by default
Specs published. Reference implementations open-sourced. No vendor lock-in.
We govern ourselves first
The ThinkNEO platform itself runs behind ThinkShield, our open defense layer — every request across the control plane is evaluated against the same enforcement engine we ship to customers. We hold our own infrastructure to the governance standard we sell, and the audit trail is the proof.
Complementary to NVIDIA Halos
Halos provides certification and inspection — assurance of how a system is built. AIRGP provides runtime decision governance — assurance of how it behaves in operation. A Halos-certified robot still needs runtime governance the moment it joins a multi-vendor fleet; a ThinkNEO-governed fleet is easier to certify, because the hash-chained audit trail is exactly the evidence stream inspectors consume.
Stop hoping your robots behave. Prove it.
Request platform access and we'll provision a sandbox fleet within 48 hours.
No sales call required. Engineering-first onboarding.