Documentation
for IT teams.

PV designs, builds and runs AI agents for businesses in the UAE. This page is for the IT and security teams who review our work: how agents connect to your systems, where data is stored, and how you stay in control.

Each project is different, so we agree the details in writing before we start. If something here doesn’t match how your organisation works, tell us and we’ll adapt.

Agents reach your systems the same way your other software does:

• APIs: the preferred route, using the system’s own API and credentials you issue for the agent.
• Webhooks: so your systems can notify the agent when something happens, such as a new invoice or a customer message.
• MCP: if a tool offers an MCP (Model Context Protocol) server, agents can use it directly, and we can build one for your in-house systems. Your own AI assistants can also reach your PV agents through MCP.
• Service accounts: a dedicated account for each agent, never a person’s login, so its access can be reviewed and revoked on its own.

We start with read-only access and add write access only for the specific actions you approve. Technical details are in the APIs Reference.

• In the UAE: agent data is stored in the UAE. Other regions are used only if you ask for them.
• Encrypted: data is encrypted at rest and in transit.
• Kept apart: each client’s agents and data are isolated from every other client’s.
• Not used for training: we never use your data to train AI models.
• Your own cloud: on request, agents can run in your organisation’s own cloud account.
• Only what’s needed: agents read the data a task requires, and we agree retention periods with you.

• Least privilege: each agent gets the smallest set of permissions its job needs.
• Human approval: sensitive steps, such as payments, sending to customers or changing records, can be set to wait for a person to approve them.
• Off switch: you can ask us to pause an agent at any time, and revoking its credentials stops it immediately.

Read more on our Security page.

Every action an agent takes is logged: what it did, when, and with which system. Your team can follow agents, raise requests and see usage in the PV Platform.

Logs can be exported on request for your own records or reviews.

• A technical contact: one person on your side who can answer questions and approve access.
• A safe place to test: a test environment or sample data, so we can build without touching live records.
• Access to the systems in scope: credentials or service accounts for the agreed systems only.

We sign an NDA before we start and are happy to complete your security questionnaire or vendor assessment.

If you need specific documents for your review, ask our team and we’ll tell you what we can provide.

• Your data stays yours: you own your data and the work your agents produce.
• Documented: we document what each agent does, which systems it uses and how it is configured.
• No lock-in on data: if you stop working with us, we return or delete your data as agreed.

Which AI models do you use?
We pick the right model for each job from leading providers such as OpenAI, Anthropic and Mistral, or open-source models. We tell you which ones a project uses.

Do you train AI models on our data?
No. Your data is used only to do your work.

Can the agents run in our own cloud?
Yes, on request. We can deploy them in your organisation’s cloud account so data never leaves it.

Do we need to install anything?
Usually not. Agents connect through the APIs, webhooks, MCP servers and accounts you give them.

What happens if an agent makes a mistake?
Every action is logged, sensitive steps can require approval, and we monitor agents after launch to catch and fix issues.

Is the service up right now?
See our status page.