How it worksProductsBlogAbout
Sign inCreate an agent

How it works

From signup to live agent in five steps.

This is the real customer path: create an account, choose the agent, checkout, watch deployment progress, then open the hosted runtime with managed Kriy.AI tokens already wired in.

Start the flowCompare agents
What the user sees after checkoutTarget under 1 min
Deploying agent

OpenClaw agent

Provisioning runtime, health checks, access URL, and Kriy.AI tokens.

74%
RuntimeOpenClaw, Hermes, or NemoClaw
DeploymentProgress, timer, health checks
AI accessManaged Kriy.AI tokens
FeedbackSupport and product loop

The first run should feel obvious.

New users should always know where they are, what is happening now, and what they can do next.

1

Create account

Sign up, verify email, and land on the agent setup path.

2

Choose agent

Choose OpenClaw for most first agents, Hermes for delegated work, or NemoClaw for secure Builder runs.

3

Checkout

Confirm the plan in Stripe. Every new paid agent includes 100,000 free Kriy.AI tokens.

4

Deploy

Watch the timer, progress bar, health checks, and token setup in the dashboard.

5

Operate

Open the agent and start work. Feedback and support stay attached to the account.

Choose by job, not by infrastructure.

The user picks the kind of work. Kriy.AI handles the VM, image, health checks, routing, and access details.

OpenClaw

Gives you a browser-first hosted agent for direct operation.

Hermes

Plans, delegates, and runs longer agent workflows.

NemoClaw

Runs secure OpenClaw experiments in controlled Builder environments.

The operating loop is part of the product.

Kriy.AI is not just a hosted shell. It connects deployment, tokens, feedback, and future memory so raw intelligence turns into disciplined action.

No provider keys in user hands

Kriy.AI token access is installed server-side. Users see status and balance instead of managing vendor secrets.

Deployment is observable

The dashboard shows deployment stage, elapsed time, target timing, and agent health instead of a silent wait.

Feedback has a destination

Support and product feedback are routed into the operating loop so failures become tracked work, not lost comments.