ChatGPT & Codex
Everythings is in ChatGPT's plugin directory, which ChatGPT shares with Codex. Install the plugin, sign in once, and ChatGPT can search, read and write your things, ask you questions on them, and draw your workspace in the chat on the web.
Install the plugin
One listing, one button, one sign-in.
- Open the Everythings plugin, or search for Everythings in ChatGPT's plugin directory.
- Press Install plugin.
- When ChatGPT connects the Everythings app, sign in on the Everythings page it opens. It lists the access ChatGPT asked for (read, or read and write); press Approve.
- Start a chat and ask: “Show me my workspaces in Everythings.”
In a ChatGPT Business or Enterprise workspace, an admin may need to allow the plugin before members can install it. The listing shows the server it connects to, https://www.everythings.app/api/mcp, so you can check you are installing the real one.
Using it in a chat
Say in Everythings once and ChatGPT knows where to look. You can also pick Everythings under the + menu in the composer. These are the starter prompts the listing suggests, with what happens.
“Show me my workspaces in Everythings.”
ChatGPT lists your workspaces and, on the web, draws your default workspace's tiles in the chat.
“Show me the questions my agents left for me in Everythings.”
ChatGPT lists the open questions across your workspaces, each with the thing it sits on and its options.
“Save this plan under my project in Everythings and ask me there whether to go ahead.”
ChatGPT creates the thing under your project and attaches a question to it. You answer from your phone or the web, and a later chat reads the answer.
“Find my packing list in Everythings.”
ChatGPT searches every workspace you can open and reads the one you mean.
“What did we write about the Kyoto itinerary?”
ChatGPT finds the thing by name and answers from its content, sub-things and comments.
People name a thing after its topic, so “add this to my books thing” or “what's on my shopping list?” is enough: ChatGPT searches by the topic word, reads the thing, and edits it in full.
What ChatGPT confirms
Read tools are annotated read-only, so ChatGPT runs them without asking. Four tools are annotated destructive, because they overwrite or delete something, and ChatGPT confirms each of them with you first.
| Action | What happens |
|---|---|
| Editing a thing | update_thing replaces the content in full, so ChatGPT reads the current version first and confirms the save. |
| Deleting a thing | delete_thing moves it to Trash, where you can restore it. |
| Renaming a workspace | rename_workspace overwrites the name. |
| Removing a mark | remove_mark deletes the mark record. |
Creating things, moving and reordering them, restoring from Trash, commenting, adding a mark and asking you a question run without a confirmation. The full list, tool by tool: Tools reference.
What you see
On the web, asking to see a workspace or a thing draws the Everythings view in the chat: a workspace as a grid of emoji tiles, a thing with its Markdown, sub-things, marks and comments, with tiles you can click.
Where the view draws
ChatGPT on the web (chatgpt.com, on a computer or in a phone browser) renders the view. The native iOS and Android apps show the same workspace or thing as text, and everything else works the same there.
Deep Research can use it
Everythings also exposes search and fetch tools in the shape ChatGPT's research mode expects, so a research task can pull from your things alongside the web.
What a click in the view does, the one thing it can write, and why it shows an agent's question read-only: The view in the chat.
Codex
The plugin directory is shared between ChatGPT and Codex, and the Everythings plugin gives both the same tools. A coding agent in Codex can read a plan from a workspace, file its results as things, and ask you a question on one of them when it needs a decision.
Starting a workspace from the Agent workflow template gives such an agent task lists, plans and an archive to work in. The Agent workflow.
Access, labels and revoking
- The plugin reaches every workspace your account can open, your own and the shared ones. The access you approved limits what it can do there.
- Everything it writes carries its label: ✦ ChatGPT on the thing and ✦ on its tile, “via ChatGPT” on a comment, and “agent” after your name on a mark. Members of a shared workspace see the same labels.
- A delete goes to Trash. ChatGPT cannot empty Trash, delete a workspace or invite anyone.
- To disconnect, revoke its row under Settings › MCP Access Tokens in Everythings: that ends every token the app holds for you, refresh tokens included, within a minute. Removing the app in ChatGPT's settings afterwards only tidies its list.
Without the directory
Where the listing is unavailable, ChatGPT can connect to the same server as a custom MCP app. That needs Developer mode, which OpenAI offers on its paid plans.
Add Everythings as a custom MCP app
- Open Settings › Security and login, find Developer mode, and turn it on.
- Open Plugins › Browse, press +, and choose Create custom MCP server, then Create MCP app.
- Fill in the form: Name Everythings, Description anything that reminds you what it is, Connection the address below, Authentication OAuth. Tick I understand and want to continue.
- Press Create, sign in to Everythings when ChatGPT sends you there, and approve the access it asks for.
https://www.everythings.app/api/mcp
ChatGPT shows the same warning for every custom app: connect only to servers you trust. This one is run by the makers of Everythings at www.everythings.app. You sign in on the Everythings page itself, so ChatGPT never sees your password, and everything above about access, labels and revoking applies the same way.
Questions
Which ChatGPT plans can install it?
Plugins are available on the plans OpenAI offers them on, and a ChatGPT workspace's admin can limit which plugins members may install. If the Install plugin button is missing or disabled, check with your workspace admin, or use the custom MCP app route below where Developer mode is available.
Why does ChatGPT ask before editing or deleting?
Everythings marks four tools as destructive under ChatGPT's definition, because they overwrite or delete something: editing a thing's content, deleting a thing (to Trash), renaming a workspace and removing a mark. ChatGPT confirms those with you each time. Reading, creating, moving, commenting and marking run without a prompt.
Does it work in the ChatGPT phone apps?
Yes, with text answers. ChatGPT's native iOS and Android apps show the same data as text where the web draws the Everythings view. chatgpt.com in a phone browser draws the view.
Is this the same as the AI Provider key in Settings?
No. That key is what AI Mode and voice inside Everythings use to call OpenAI. The plugin works the other way round: ChatGPT reaches into your Everythings data, and needs no key.
Can I use Claude and ChatGPT at the same time?
Yes. Each connects on its own and each labels its work with its own name, so a thing reads “✦ ChatGPT” or “✦ Claude”. Both are listed, and can be revoked separately, under Settings › MCP Access Tokens.
What does ChatGPT send to Everythings?
The tool calls it makes on your behalf: a search query, a thing to read, content to save. ChatGPT shows its own standard notice that it may share relevant chat context with a connected app; Everythings stores only what a tool call writes into a thing, a comment or a mark, and the privacy policy describes the rest.