Frequently asked questions
Straight answers about credits, data, models, infrastructure and contracting.
- Does Deep Agents charge per token?
- No. You subscribe to a monthly plan with a number of credits included. Each run consumes credits from the company balance, calculated from the real cost of the model used — lighter models consume far less per answer. Balance and consumption appear in real time in the workspace. See the plans and the credits included
- How do credits work?
- Each plan allocates credits per month. If the operation goes past that allocation, extra packs can be bought inside the platform: they never expire and are only consumed once the plan's credits run out. See the credit packs
- Can I run Deep Agents on my own infrastructure?
- Yes. Deployment can be on-premises, on your own hardware and air-gapped if needed; in a private cloud (VPC), with an isolated perimeter; or in the cloud the organization already uses — AWS, Azure, GCP or a national provider. See the deployment options
- Are my documents copied outside the organization?
- No. Agents work directly on the organization's databases, documents and internal systems, and nothing is copied into third-party tools. Sensitivity analysis decides what may be processed outside national infrastructure; everything else stays in your environment, with the source of each answer on record. See how data stays in your territory
- What about LGPD, Brazil's data protection law?
- Deep Agents provides the controls an AI operation needs in order to handle personal data: the data stays in the environment you choose, sensitivity analysis decides what leaves it, access follows the permissions and policies your team defines, and every run is recorded with source, user, model and cost. Compliance also depends on how the organization uses the platform. Read the privacy policy
- Which AI models can I use?
- Routing covers the leading providers — GPT-5, Claude, Gemini, Llama, Qwen, DeepSeek and NVIDIA — plus models running on your own infrastructure. The available set depends on the plan you are on. See which models each plan includes
- Can I audit what the AI did?
- Yes. The audit trail records every run: the time, what was executed, the responsible user, the model used, the cost and the result. The record can be exported as CSV. See per-run auditing
- Where do agents become available to people?
- A published agent can be opened on a page of its own, embedded in the organization's website or consumed directly through the API. Read the Agent API documentation
- Which languages does the platform support?
- The Deep Agents interface is available in Portuguese, English and Spanish.
- How do I get started?
- You can create an account on the free plan and launch your first agent, or talk to the team about a fixed-scope deployment: integration with your systems, access setup, first agents and team training. Talk to the team on WhatsApp