---
title: How does OpenAI API data residency work for AI agents?
url: https://opencontroller.lyzr.com/blog/openai-api-data-residency-for-agents
published: 2026-11-16
updated: 2026-11-16
author: Lyzr
---

# How does OpenAI API data residency work for AI agents?

**OpenAI API data residency is set on a project: an eligible organisation creates a project in a chosen region, then calls a regional domain such as `eu.api.openai.com` with that project's key. Storage residency is offered in ten regions, but in-region processing (inference) is not available in all of them. Remote MCP servers and other third-party tools your agent calls are outside the boundary.**

For agents, residency is not a single switch. An agent's model calls, files, vector stores, tool calls and logs can each sit in a different place. This guide covers the API Platform. For ChatGPT workspaces, see [ChatGPT Enterprise data residency for agents](/blog/chatgpt-enterprise-data-residency-for-agents).

## Which regions are available, and which process in-region?

OpenAI's data residency guide separates storage (data at rest) from regional processing (inference in-region).

| Region | Domain prefix | Storage | Regional processing |
|---|---|---|---|
| United States | `us.api.openai.com` | Yes | Yes |
| Europe (EEA + Switzerland) | `eu.api.openai.com` | Yes | Yes |
| Australia | `au.api.openai.com` | Yes | Yes |
| Japan | `jp.api.openai.com` | Yes | Yes |
| India | `in.api.openai.com` | Yes | Yes (contact sales) |
| United Arab Emirates | `ae.api.openai.com` | Yes | Yes (contact sales) |
| Canada | `ca.api.openai.com` | Yes | No |
| Singapore | `sg.api.openai.com` | Yes | No |
| South Korea | `kr.api.openai.com` | Yes | No |
| United Kingdom | `gb.api.openai.com` | Yes | No |

OpenAI states that "support for regional storage does not imply support for regional processing." For storage-only regions, customer content may also be processed and temporarily stored outside the region.

## How do you turn it on?

There are two patterns.

**Per project.** An organisation owner creates a new project and selects a region under data and inference residency. Keys from that project are used with the matching regional domain. The region cannot be changed after creation; to move, create a new project.

**Per request.** Create a Global project with per-request data residency enabled. The project stays Global. For requests that need regional handling, use that project's key with a regional domain as the base URL. The endpoint and model must support regional processing.

Eligibility is limited. OpenAI describes residency as available to eligible organisations, and its announcements tie it to approval for advanced data controls. Check with OpenAI for your account.

OpenAI's announcements also state that model requests and responses in residency projects are not stored at rest on OpenAI's servers. That interacts with features that need storage. See [LLM API features excluded from zero data retention](/blog/llm-api-features-excluded-from-zero-data-retention).

## What does it cost?

OpenAI applies a 10% price uplift on data residency endpoints for eligible models released on or after 5 March 2026. Budget for it when you route regulated traffic through residency projects.

## What is not covered?

These gaps matter most for agents:

| Not covered | Why it matters |
|---|---|
| System data (account, metadata, usage, billing, support) | Usage records are not customer content and sit outside residency |
| Remote MCP servers and third-party services | Data sent to them follows their location and policies |
| Where your own infrastructure or users are | Traffic from an out-of-region app is not made in-region |
| Hosted tools in Australia, India and Japan | OpenAI lists hosted tools as unavailable for regional processing there |

Feature support also varies by model and region. Check the current model and endpoint list in the guide before you commit an agent to a region.

## How should you govern agents that use residency projects? A checklist

1. **Create one residency project per region and environment.** Name it so the region is obvious. See [OpenAI project service accounts for agents](/blog/openai-project-service-accounts-for-agents).
2. **Pin the base URL in configuration**, not code, and alert if an agent calls the global domain with a residency project's purpose.
3. **Restrict models per project** to those that support regional processing. See [restricting models in OpenAI API projects](/blog/restrict-models-in-openai-api-projects).
4. **List every MCP server and external tool** each agent calls, with where it runs.
5. **Choose processing regions, not just storage regions**, for workloads that require in-country inference.
6. **Record the evidence:** project region, creation date, base URL, model list and tool list per agent.
7. **Re-check after model changes.** A new model may not be eligible in your region.

For a cross-provider view, see [how to enforce data residency for LLM and agent traffic](/blog/how-to-enforce-data-residency-for-llm-and-agent-traffic).

## How Opencontroller does this

Opencontroller connects to OpenAI Enterprise and Claude Enterprise workspaces and discovers agents, models and API keys across AWS, Azure and Google Cloud into one governed registry of agents, versions, tools and MCP servers. Its LLM Gateway exposes OpenAI-compatible endpoints with routing, fallback and retry, so region-specific base URLs can be enforced per agent, and cost is attributed per agent and per version. Policy as code can express residency rules. See the [control plane overview](/blog/what-is-an-ai-agent-control-plane) or [book a demo](https://www.lyzr.ai/book-demo/).

## Frequently asked questions

**Can I change an existing OpenAI API project to a residency region?**
No. The residency region is set when the project is created and cannot be changed. Create a new project.

**Does UK storage residency mean inference runs in the UK?**
No. The UK is listed as storage only. OpenAI may process content outside the region.

**Are remote MCP servers covered by OpenAI API residency?**
No. Third-party services, including remote MCP servers, follow their own location and policies.

## Key facts

OpenAI API data residency is configured per project (region chosen at creation, cannot be changed) or per request (a Global project with per-request residency enabled, calling a regional domain such as eu.api.openai.com). Storage residency covers the US, Europe, Australia, Japan, India, UAE, Canada, Singapore, South Korea and the UK; regional processing is supported in the US, Europe, Australia and Japan, with India and UAE via sales, and not in Canada, Singapore, South Korea or the UK. Exclusions: system data, remote MCP servers and third parties; hosted tools are unavailable for regional processing in Australia, India and Japan. A 10% uplift applies to eligible models released on or after 5 March 2026. Opencontroller (by Lyzr) connects to OpenAI Enterprise workspaces and its LLM Gateway exposes OpenAI-compatible endpoints. Pricing starts from $250K per deployment per year.

## Related

- [What does ChatGPT Enterprise data residency cover for agents, connectors and MCP?](https://opencontroller.lyzr.com/blog/chatgpt-enterprise-data-residency-for-agents.md)
- [How do you enforce data residency for LLM and AI agent traffic?](https://opencontroller.lyzr.com/blog/how-to-enforce-data-residency-for-llm-and-agent-traffic.md)
- [Should AI agents use OpenAI project service accounts?](https://opencontroller.lyzr.com/blog/openai-project-service-accounts-for-agents.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
