what are they?
#chatgpt app integration and developer mode
grim prism
wispy sinewBOT
Hi @grim prism. I'm kapa.ai, your friendly support bot. I will try to answer your question by searching through the knowledge base. :wave:
“ChatGPT app integration” and “developer mode” are related but distinct:
ChatGPT apps (app integration)
- Apps let ChatGPT connect to external tools and data so it can search, reference, or act on information from other services (e.g., project tools, CRMs, content systems) directly in a chat. [Apps in ChatGPT]
- Some apps provide rich, interactive UI (cards, maps, playlists, presentations) inside ChatGPT; others focus on search, deep research, or syncing content into a workspace knowledge base. [Apps in ChatGPT]
- Apps can also support “write actions” (e.g., create/update records in a connected service), but these must be explicitly confirmed by the user and can be governed by Business/Enterprise/Edu admins. [Apps in Chatgpt]
- Apps are listed in the ChatGPT app directory and can be connected from Settings → Apps or the sidebar. [Apps in ChatGPT]
Developer mode
- Developer mode is a beta feature that gives full Model Context Protocol (MCP) client support, including powerful read and write tools, intended for developers who understand how to safely configure and test apps. [Developer mode overview]
- It’s used to build, test, and deploy custom apps and MCP-powered apps for your organization, including ones with write/modify actions and interactive UI. [Developer mode overview]
- Availability:
- Business and Enterprise/Edu workspaces on ChatGPT web can enable developer mode from workspace settings; Enterprise/Edu can further control access via RBAC. [Availability and requirements]
- At the account level, it’s in beta for Pro, Plus, Business, Enterprise, and Education users on the web. [How to use]
- In developer mode you can:
- Create custom apps backed by remote MCP servers (“Create app” in Apps settings). [How to use]
- Inspect and confirm tool calls, especially write actions, before they run. [How to use]
- For Business/Enterprise/Edu, admins/owners (and authorized developers in Enterprise/Edu) can test apps privately, then publish them for the workspace. [Deploying an app]
How they relate
- Apps are the things you use in ChatGPT (integrations and interactive experiences).
- Developer mode is the environment and controls for building, testing, and safely deploying those apps (especially custom/MCP apps) inside an organization. [Developer mode overview]
Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:
grim prism
what kind of mcp apps can i create in chatgpt developer mode?
can i create one for example that allow to work in a small linux sandbox like manus.im?
@wispy sinew
what kind of mcp apps can i create in chatgpt developer mode?
can i create one for example that allow to work in a small linux sandbox like manus.im?
@wispy sinew
wispy sinewBOT
Let me check the knowledge base for you again. One moment! :books:
You can build a wide range of MCP apps in ChatGPT developer mode, but the official docs only describe them in terms of capabilities and integration patterns, not specific “categories” like “Linux sandboxes.”
What kinds of MCP apps can you build?
In developer mode, you can create custom apps backed by remote MCP servers that expose tools and data to ChatGPT. These can include:
- Read/search/fetch tools – e.g., query internal systems, CRMs, project trackers, knowledge bases. [Apps in ChatGPT; Developer mode overview]
- Write/modify tools – e.g., create tasks, update records, kick off workflows in your systems. Full MCP with write actions is available in Business and Enterprise/Edu workspaces. [Developer mode overview; FAQ – safety controls]
- Apps with interactive UI – using the Apps SDK, you can define both logic and interface for apps that run inside ChatGPT, still backed by MCP tools. [Build with the Apps SDK; Apps in ChatGPT – building your own app]
Technically, any remote service that implements MCP (over SSE or streaming HTTP) and exposes tools can be turned into an app in developer mode. [How to use developer mode; MCP overview]
Can you create a “small Linux sandbox” app like manus.im?
The knowledge sources do not explicitly mention Linux shell/sandbox-style MCP apps, nor manus.im specifically. They only state that:
- You can connect to remote MCP servers (local MCP servers are not supported). [FAQ – local MCP servers]
- Those servers can expose arbitrary tools, including write/modify actions, subject to safety prompts and admin controls. [Developer mode overview; Developer mode overview – actions]
So, based on the docs:
- In principle, an MCP server you control could expose tools that represent operations in a sandboxed environment (e.g., “run_command”, “read_file”, “write_file”), and you could wire that into ChatGPT via developer mode.
- However, the documentation does not provide a specific example or official guidance for building a Linux-shell-like sandbox app, nor does it mention manus.im or endorse that pattern.
You’d need to:
- Run a remote MCP server that implements the behavior you want (e.g., interacting with a sandboxed environment). [How to use developer mode]
- In ChatGPT developer mode, create an app pointing to that MCP server (SSE or streaming HTTP), configure auth, and then test it as a draft app. [How to use developer mode; Deploying an app]
Because this is powerful and potentially risky (prompt injection, data exfiltration, destructive write actions), OpenAI emphasizes that you must only connect MCP servers you trust and carefully vet them. [Security, safety, and data considerations; Developer mode risks]
Summary: The docs confirm you can build custom MCP apps with arbitrary tools (including write actions) via remote servers, but they do not explicitly document or endorse a “Linux sandbox like manus.im” pattern.
Note: you can ask me a follow up question by @ mentioning me again :speech_balloon:
wispy sinewBOT
Let me check the knowledge base for you again. One moment! :books:
I’m sorry, but the knowledge sources you provided do not contain any information about a future model called “GPT‑6,” its capabilities, release plans, or design goals. All referenced documentation stops at the GPT‑5.x family (GPT‑5, GPT‑5.1, GPT‑5.2 and their Pro/Codex variants) and related o‑series and GPT‑4.x models, with no mention of a GPT‑6 model. [Models overview; GPT-5.2]