Frameworks and models

Your AI employees do not have to be ours.

Run OpenClaw or an agent your team built, on the model you choose, with the same controls as every other employee.

Pay for models with your own keys or with Humanik credit. The systems you connected stay connected when the agent changes.
  • OpenClaw oryour own agent
  • A model foreach employee
  • Your keys orHumanik credit
  • Budgets onboth

Frameworks

Run the agent you want.

The framework is the agent program inside an employee: the part that decides what to do and calls tools. Run OpenClaw, the open source agent, or an agent your team built.

Either way it runs on HumanikOS with the same permissions, logs, compute settings and budgets as every other employee. Two employees in one workspace can run different agents at the same time.

Front desk · Agents

The same controls

Every agent gets the same place to work.

The platform builds the employee's environment and then starts the agent inside it. So whichever agent you run, it gets the same machine, rules and settings. One control per slide.

Every agent gets

A machine of its own

Each employee runs on its own machine, with a filesystem and a terminal, in Humanik Cloud or on a computer you own.

Connections

Your connections move with you.

Each system you connect is stored as data: where it lives, how it signs in, which key in the vault it uses, and its endpoints. When an employee starts, that record is turned into the tools and instructions its agent reads.

So nothing you connected is tied to one agent. Change the agent, and the connections are still there.

How connections work

Models

Pick the model for each employee.

Models · Sales workspace

Give each employee the model that suits its work: a small, quick model for sorting messages, the strongest one for open questions. Choose from hosted models, or run a model on a computer you own, where there is no charge per word.

In Nova, each person picks their own model, and it applies to their next message. An employee reads its model when it starts, so a change applies at its next start.

Run models on your own computers

For admins

Use your own keys, or ours.

The account owner adds the company's model keys once, for Anthropic and for OpenRouter, and gives each one to the workspaces that may use it. Then each employee can be set to use a key. The people who set an employee to use a key never see it. Only its last four characters are shown.

Use on your own key is billed by your provider, at your rates. Or skip the keys and pay with Humanik credit.

One thing to know

Nova starts on Auto, and Auto always uses Humanik credit, even when a key is set up. Pick a model to use your own key.

What Humanik credit costs

Budgets

Set budgets that count your own keys too.

Set a limit on model spend for the whole account, one workspace or one employee, by the day, week or month. Choose what it counts: Humanik credit, your own keys, or both. At the limit, it stops the work, warns, or only sends an alert.

We are not paid for use on your own keys, and a budget can still limit it. A team lead can set their own team's budget without seeing the account's invoices. Budgets cover model spend only. Compute depends on the size of each machine and how long it stays awake.

See how pricing works

Why this is possible

The agent is a part, not the product.

Many AI products are built around their own agent, so the agent cannot be replaced. HumanikOS is built the other way round. We build everything around the agent: the machine, the identity, the vault, the data, the schedules and the billing. The agent is one part, and you choose it.

How HumanikOS fits together

Bring the agent. Keep the controls.

Start on OpenClaw and Humanik credit. Add your own agent, your own keys or your own computer when you are ready, one employee at a time.