---
title: What does ChatGPT Enterprise data residency cover for agents, connectors and MCP?
url: https://opencontroller.lyzr.com/blog/chatgpt-enterprise-data-residency-for-agents
published: 2026-11-16
updated: 2026-11-16
author: Lyzr
---

# What does ChatGPT Enterprise data residency cover for agents, connectors and MCP?

**ChatGPT Enterprise data residency keeps conversations, files, memory, custom GPTs and similar content at rest in your chosen region. It does not cover data processed through Apps, MCP connectors or web search, which may leave the region. Inference residency, where the model runs in-region, is available only in Europe, the US and the UAE.**

That gap matters most for agents. An agent that calls an MCP server or searches the web moves data outside the residency boundary by design. This guide shows what is covered and how to govern the rest.

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

## Which regions are available?

OpenAI separates storage from processing.

| Type | What it means | Regions |
|---|---|---|
| Data residency | Content stored at rest in-region | Australia, Canada, Europe (EEA + Switzerland), India, Japan, Singapore, South Korea, UAE, UK, US |
| Inference residency | GPU inference runs in-region | Europe (EEA + Switzerland), US, UAE |

Rules to note:

- Data residency is offered to **new** ChatGPT Enterprise and Edu workspaces, at no additional cost.
- Existing workspaces are not migrated automatically.
- Inference residency requires data residency in the same region.
- Changes to inference residency apply to new conversations only.
- Some features are limited in UAE inference residency, including image generation and internal search.

## What data is covered?

OpenAI lists these categories as stored in-region:

1. Conversations (text, image, voice)
2. Files
3. ChatGPT Memory
4. Custom GPTs and their prompts and outputs
5. Code Interpreter and Data Analysis artifacts
6. Image generation inputs and outputs
7. Backups and replicas of that content

Inference residency covers the same categories, including embeddings derived from them.

## What is not covered?

This is the list agent owners need:

| Not covered | Why it matters for agents |
|---|---|
| Data processed via Apps and MCP | Connector calls send data to the connected service, wherever it runs |
| Web search (if enabled) | Queries leave the boundary |
| Transient processing steps | Short-lived processing may occur elsewhere |
| Workspace metadata, billing, user logins | Identity and admin data |
| CPU processing, auth, routing, analytics | Even with inference residency |

Unsupported in residency workspaces:

- Third-party GPTs
- Codex Web (the Codex app and CLI are supported)
- Features not generally available (alpha or beta)
- Apple Intelligence integrations

Residency workspaces also cannot set GPT sharing to "Anyone". Shared GPTs work only for users in the same residency region.

## How should you govern agents and connectors in a residency workspace?

1. **Inventory every connector and MCP server** enabled in the workspace. See [controlling ChatGPT Enterprise connectors and MCP](/blog/control-chatgpt-enterprise-connectors-and-mcp).
2. **Record where each connected service stores and processes data.** A connector to an EU-hosted system keeps data in the EU; one to a US SaaS does not.
3. **Classify connectors by residency risk**: in-region, out-of-region, unknown.
4. **Disable out-of-region connectors** for users or GPTs handling regulated data.
5. **Decide on web search.** Turn it off for workspaces or roles where queries must stay in-region.
6. **Block third-party GPTs** and review custom GPTs and agents. See [governing GPTs and workspace agents](/blog/govern-chatgpt-enterprise-gpts-and-agents).
7. **Check new features before enabling.** Beta features fall outside residency.
8. **Document the boundary** in your DPIA and records of processing.

## Does residency make ChatGPT agents compliant?

Residency is evidence for data location controls, not a compliance outcome on its own. Regulators and auditors will also ask:

- which data flows leave the region and why,
- what contracts cover those flows,
- who approved each connector,
- and how you would know if a new one appeared.

The answer to the last question is continuous discovery, not a one-time review.

## How Opencontroller does this

Opencontroller connects to OpenAI Enterprise and Claude Enterprise workspaces, so their agents and usage appear in one governed registry with agents from AWS, Azure, Google Cloud and SaaS platforms. The registry tracks agents, versions, tools and MCP servers. Posture flags risky configuration, and policy as code can express residency rules. Its mappings to NIST, ISO 42001, OWASP and MITRE ATLAS help assemble evidence. [Book a demo](https://www.lyzr.ai/book-demo/) to see workspace discovery.

## Frequently asked questions

**Is inference residency available in every residency region?**
No. Inference residency is available in Europe, the US and the UAE. Other regions offer storage at rest only.

**Are custom GPTs covered by residency?**
Yes, custom GPTs and their prompts and outputs are covered. Third-party GPTs are not supported in residency workspaces.

**Does residency apply to ChatGPT Sites?**
For US or EU workspaces, new Sites use the workspace region for storage and hosting. In other regions the regional setting does not apply to Sites.

## Key facts

OpenAI offers ChatGPT data residency (storage at rest) in Australia, Canada, Europe (EEA and Switzerland), India, Japan, Singapore, South Korea, UAE, UK and US for new Enterprise and Edu workspaces, at no extra cost. Inference residency is available only in Europe, the US and the UAE, and requires data residency in the same region. Covered data includes conversations, files, memory, custom GPTs, Code Interpreter artifacts and image generation. Data processed through external integrations such as Apps and MCP, or web search, is not covered and may leave the region. Third-party GPTs and Codex Web are unsupported. Residency-enabled workspaces cannot share GPTs with Anyone. Existing workspaces are not migrated automatically. Opencontroller connects to OpenAI Enterprise workspaces for discovery. Opencontroller pricing starts from $250K per deployment per year.

## Related

- [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)
- [How do admins control connectors and custom MCP servers in ChatGPT Enterprise?](https://opencontroller.lyzr.com/blog/control-chatgpt-enterprise-connectors-and-mcp.md)
- [How do you govern GPTs and workspace agents in ChatGPT Enterprise?](https://opencontroller.lyzr.com/blog/govern-chatgpt-enterprise-gpts-and-agents.md)
- [More on Platform Guides](https://opencontroller.lyzr.com/blog/cluster/platform-guides)
