ChatGPT Projects API: App Projects versus API Projects

The phrase "ChatGPT Projects API" means two different things depending on context. ChatGPT Projects are a conversation-organization feature in the ChatGPT app; API Projects are an organizational unit in the OpenAI developer platform. They operate at different layers and serve different users.

Understanding the difference matters if you are building on the OpenAI API, managing team access to ChatGPT, or trying to keep project context available to an AI assistant. Tools like fenbs address a third layer: a shared work record that any connected assistant can read and update, independent of which project or API key was used to access the model.

ChatGPT Projects in the ChatGPT app

ChatGPT Projects are a feature in ChatGPT Work that lets you group conversations around a topic or body of work. Each project can have its own instructions, memory settings, connected files, and data sources. When you open a conversation in a project, ChatGPT carries those settings into the session.

This is primarily a context management feature for people using ChatGPT day-to-day. A project for client work might include a brief, the client's style guide and a list of standing instructions; a project for a software product might include architecture notes and coding conventions.

ChatGPT Projects do not expose a public API of their own—they are a user-interface layer, not an integration surface. You cannot query a ChatGPT Project from an external application the same way you query the OpenAI API. For a comparison of how ChatGPT and Claude organize project context differently, see Claude projects vs ChatGPT projects.

Projects in the OpenAI API platform

OpenAI API Projects are an organizational unit in the OpenAI developer platform at platform.openai.com. They exist for account holders who build with the API: developers, teams, and companies running applications on top of OpenAI's models.

An API Project groups API keys, usage limits, spending data, and fine-tuned models together. If you are building a product that calls the OpenAI API, you would create an API Project for it, give each deployment its own key within that project, and track usage per project rather than mixing all calls under one account.

• Each API Project has its own API keys, separate from other projects in the organization.
• Usage and spending are reported per project, making cost tracking cleaner.
• Access permissions control which team members can manage the project, create keys, or view usage.
• API Projects are a billing and access-management feature, not a place where conversation context lives.

How the two kinds of project interact

A ChatGPT app Project and an OpenAI API Project are not the same thing and do not directly interact. A ChatGPT user who creates a "Client Work" project in the ChatGPT app is not creating an API Project; they are organizing their conversations. A developer who creates an API Project is managing API access, not creating a ChatGPT conversation folder.

The confusion arises because both use the word "project." The simplest distinction: if you are using chatgpt.com in a browser or the ChatGPT app, you are working with ChatGPT Projects. If you are at platform.openai.com building or managing API access, you are working with API Projects.

If your team builds a product on the OpenAI API and also uses ChatGPT Work for internal tasks, you may have both. They do not share configuration or conversation history. See ChatGPT connectors and apps for how external tools connect to ChatGPT at the app level.

Context boundaries in ChatGPT Projects

Context stored in a ChatGPT Project—its instructions, connected files, and memory—stays in that project and in ChatGPT. It is not visible to other AI apps, to teammates who are not in the same workspace, or to applications calling the OpenAI API directly.

If multiple people on a team use ChatGPT Work and you want shared context between them, that context needs to be in a shared project or in an external tool that everyone connects to. A shared ChatGPT Project in a ChatGPT Work workspace lets team members start from the same instructions and files. An external shared record—one that also works with other AI apps—can go further.

Accessing project context through the OpenAI API

If you are building an application that calls the OpenAI API, there is no endpoint that returns the contents of a ChatGPT Work Project. API calls go to the model, not to a user's conversation history or project context. Any context your application needs must be supplied in the prompt or in the system message.

For developers building context-aware applications on the OpenAI API, the pattern is to supply relevant context programmatically: retrieve it from a work record or knowledge store and include it in each request. That is the role a tool like fenbs plays for AI apps that connect to it over MCP.

How to do this with fenbs

1. Create a free account and add the project to your fenbs board. Write a clear description and list the first tasks.
2. Record the standing rules in Decisions and rules—what applies to every session in this project, with the person who decided each one.
3. Connect ChatGPT to fenbs via MCP following the developer-mode instructions at Integrations.
4. At the start of each ChatGPT Work session, ask the assistant to call the context tool. It returns Rules, Where we left off and what is In progress.
5. Before the session ends, save progress in fenbs. The next session—in ChatGPT, Claude, or any other connected app—starts from an accurate record.