OneWare Agents

OneWare Agents give you direct access to GPT, Claude, DeepSeek, and other leading AI models inside OneWare Studio. The models are connected to OneWare Studio and ONE AI, so they can use the software for you to build your ONE AI models: you describe what you want to detect or classify, and the agent sets up the project, prepares the dataset, trains, tests and exports the model, while you stay in control of every step.

Unlike the Chat Assistant on the OneWare Cloud website, which only answers questions, OneWare Agents work directly in your projects (comparison).

OneWare Agents were previously called OneWare Cloud AI.

What it can do with ONE AI

• Plan your project: choose the right task (classification, object detection, segmentation), label design, input resolution and settings that fit your target hardware.
• Create and configure projects: create a ONE AI project, add labels, prefilters and augmentations, set the annotation and capability mode and adjust model settings.
• Build your dataset: check your dataset for problems, capture images with a camera script, pre-label real images with SAM, or generate a synthetic dataset when you have little or no real data.
• Train, test and export: start trainings in OneWare Cloud, follow the progress, read the training logs and explain failed runs, start tests and exports and download the ONNX model.
• Verify and demonstrate: run the exported model on held-out images and create interactive scripts that show the trained model live.
• Answer questions about ONE AI, your project and your training results.

Requirements

• OneWare Studio with the latest ONE AI extension.
• A OneWare account. The agent uses your active organization (Organizations).
• The organization is on the Pro Trial or a Pro plan. The agent is not included in the Free plan (Plans & Credits).
• Enough Compute Credits, and the owners or admins of the organization have not disabled OneWare Agents under Settings → Policies.

Get started

1. Open the chat in OneWare Studio under View → Tool Windows → AI Chat.
2. Select OneWare Cloud as provider in the chat header.
3. If you are not signed in yet, open the menu in the chat header and click Sign in to OneWare Cloud.
4. Choose a model at the bottom of the chat. Hover over a model to see its details.
5. Choose the mode Agent and describe your task, for example:

   Create an object detection project that finds scratches on metal parts. My images are in C:\data\scratches. Train a model that runs on a Raspberry Pi and export it as ONNX.

The agent works in the currently open ONE AI project and asks you before it runs actions that need your approval.

Modes

Mode	What it does
Agent	Full access: researches, edits files and runs tools to complete the task
Plan	Researches your project and works out a plan before anything is changed
Ask	Answers questions without changing anything

In the same list you also find ONE AI specialists that focus on one part of the workflow: ONE AI Project Architect, ONE AI Dataset Specialist, ONE AI Training Specialist and ONE AI Automation Specialist.

Approvals

The approval mode at the bottom of the chat decides how much the agent may do on its own:

• Default: the agent asks before it runs tools that need confirmation, for example editing files outside your project or running terminal commands.
• Bypass Approval: all actions are approved automatically.
• Autopilot: like Bypass Approval, and the agent also answers its own questions.

Start with Default until you are familiar with the agent.

Costs

The agent is billed in Compute Credits of your active organization, using the included credits of the plan first:

• Chat: per token of the selected model. Every member can see the models and their prices under Settings → Policies of the organization. The chat shows the tokens of the last request and of the session.
• Trainings that the agent starts in OneWare Cloud cost the same as trainings you start yourself (50 Compute Credits per minute). Trainings on a self-hosted worker are free of charge.

Owners and admins can see the usage per member in the Usage tab of the organization and set monthly budgets for members.

Models and privacy

• Each model is marked with where it is processed: EU models are processed only in data centers inside the European Union, Global models can be processed worldwide, including the USA.
• Owners and admins can allow all models, EU models only, or disable OneWare Agents under Settings → Policies.
• OneWare Cloud does not store your prompts or the answers. The chat history stays on your computer. Only usage (model, tokens, credits) is stored for billing.

Do not enter personal data in the chat. See the privacy policy and the Terms of Service (clause 3.15) for details.

OneWare Agents vs. Chat Assistant

The Chat Assistant answers questions. OneWare Agents do the work for you:

	Chat Assistant	OneWare Agents
Where	OneWare Cloud website	Chat in OneWare Studio
Answers questions about ONE AI	✓	✓
Works directly in your local project	–	✓
Creates and changes projects, datasets, settings	–	✓
Starts trainings, tests and exports	–	✓
Runs tools and scripts in OneWare Studio	–	✓
Choice of language models (EU or Global)	–	✓
Plans	All plans	Pro Trial and Pro plans