Robot Onboarding

This guide is organized around one outcome: make the robot work. Follow these 6 steps in order and you should end with a robot that is online, conversational, able to perform one real task, and safe to hand over to operations.

Time estimate

First-time onboarding takes about 15–30 minutes. If model templates and agent configs are already in place, subsequent devices can be registered in under 5 minutes. If you are deploying robots at scale, we recommend using the CLI to automate registration and configuration sync.

What “working” means

Before handoff, confirm these four outcomes:

• The robot stays online and keeps syncing configuration
• It can hold a basic conversation for the target use case
• It can complete at least one core task
• Operators can inspect and take over through Teleops

Step 1: Prepare a Robot Model Template

Why models? A Robot Model is a hardware “template” that defines the 3D model, sensor configuration, and default parameters. When registering a new robot, you select a model and automatically inherit all its settings — no need to configure each device from scratch.

Model Management: Unified templates for different hardware specs

How to

1. Go to Platform → Robot Models
2. Click Create Model and fill in:
   • Name — e.g., “Patrol Dog Pro” or “Reception G1”
   • Manufacturer — Hardware vendor info
   • Description — Intended use case and capabilities
3. Upload a URDF/MJCF model file (optional but recommended):
   • Package the .urdf file and mesh resources into a .zip
   • The system auto-parses and provides a 3D preview
   • Use the Diagnostics feature to check model integrity

Note

If you don’t have a URDF file yet, skip the model upload for now. Core functionality doesn’t depend on the 3D model.

See also: Robot Models reference

Step 2: Register the Robot

Create a new device from the Dashboard or Robot List page.

Registering a new device: Name and select a model template

How to

1. Click Add Robot
2. Select an existing Robot Model (or use a custom type)
3. Fill in device info:
   • Name — We recommend a “location + role + number” convention, e.g., “Mall-Greeter-01”
   • Description — Responsibilities and deployment location

Naming Conventions

Good naming pays off as your fleet scales:

Pattern	Example	Best for
Location + Role + Number	WestField-Greeter-01	Retail, hospitality
Project + Number	Apollo-Proto-03	R&D and testing
Building + Floor + Number	A3-12F-Security-02	Property and security

Step 3: Install Ticos Agent

After registration, you need to install and start the client program on the physical hardware side.

Robot Detail: Click Connect to get the installation command

Ticos Agent is the core bridge program deployed on robot devices (Open Source: tiwater/ticos-agent). It connects the physical device with the Ticos Cloud Server (api.ticos.ai) and is responsible for:

• Maintaining a persistent connection with the cloud hub
• Automatically pulling the latest model, agent, and scene configurations
• Real-time processing of local perception inputs and triggering physical actions
• Bridging hardware drivers with Ticos remote control and Stardust services

How to

1. Log in to the robot’s onboard computing platform (e.g., Jetson, Raspberry Pi, or x86 IPC).
2. Click Connect in the Ticos Web console to get the one-click installation command for this robot.
3. Run the installation script in the robot terminal (e.g., curl -sSL https://get.ticos.ai/install.sh | bash -s -- --token <Token>).
4. Wait for the terminal to indicate a successful connection, then return to the Ticos Web dashboard and confirm the device status has changed to Online.

Step 4: Make the Robot Converse Reliably

The Agent is the robot’s “AI brain.” Do not try to configure everything at once. First make the robot reliably greet, understand, and answer inside a safe boundary.

Agent Configuration: Set model, personality, and skills

Minimum viable setup

Setting	Priority	Description
Model	Required	Choose an LLM (GPT-4o, Claude, DeepSeek, etc.)
Personality/Prompt	Required	Define role, tone, behavioral rules, and boundaries
Voice	Recommended	Select TTS voice for spoken responses
Skills (MCP Tools)	Recommended	Mount the tools the robot needs
Knowledge Base	As needed	Upload domain documents for RAG retrieval

Tip

For first-time onboarding, start with Model + Personality — the two required settings. This ensures the robot can hold a conversation. Voice, skills, and knowledge can be added incrementally after verification.

Continue here: Make the Robot Converse

Step 5: Teach the Robot One Job

The goal here is not to fill every page in the product. The goal is to close one real workflow. In practice that usually means two things:

• Task knowledge and business flow: what to say and what steps to follow
• Scene and navigation: where to go, which routes to use, and what zones matter

Expertise Configuration: Manage scenes, knowledge, and business flows

Recommended order

1. Start in the robot detail page’s Expertise area and add the minimum knowledge, scripts, role, or flow the robot needs
2. If the robot moves in space, go to Expertise → Scenes and bind a map plus routes
3. Configure scene elements:
   • Map — 2D/3D environment map data
   • Routes — Predefined patrol or service paths
   • Zones — Functional areas (charging, restricted, service zones)

Note

If your robot doesn’t require navigation (e.g., a fixed-position receptionist), you can skip scene configuration.

Continue here:

• Teach Tasks and Knowledge
• Bind Scene and Navigation

Step 6: Go-Live Verification

Before handing the robot to the operations team, run a complete end-to-end check.

Teleops: Perform final video and movement verification before go-live

Verification Checklist

Connectivity

• Robot shows as Online in the dashboard
• Health metrics (CPU, memory, network) show no alerts

Remote Control

• Open the Teleops — video stream loads correctly
• Sensor panels refresh with live data
• Keyboard commands produce correct robot movement

Agent Conversation

• Trigger a test conversation — agent responds normally
• Response style matches personality config (tone, role)
• Knowledge base queries hit expected document fragments (if RAG is configured)

Scene & Navigation

• Scene map loads correctly (if configured)
• Navigation commands execute as expected (if applicable)

Common Issues

Symptom	What to check
Robot shows offline	Verify device network and identifier match
No video in Teleops	Confirm media service is configured; check WebRTC channel
Agent doesn’t respond	Verify model service API key is valid
Hallucinations in responses	Refine system prompt boundaries; add knowledge base content

Onboarding Complete

Once verification passes, the robot is ready for production handoff. Next steps:

• Daily Operations — Monitoring, control, and post-incident review
• Platform Assets — Shared setup, knowledge, identity assets, and team governance
• Tools — MCP tool integration and real-time diagnostics