Claude
Everythings is published in Claude's connector directory. Connect it once and Claude can read and write your workspaces on the web, in Claude Desktop and in Cowork, and draw them in the chat as emoji tiles you can click.
Connect from the directory
There is nothing to paste. Open the listing and click Connect.
- Open claude.ai/directory/everythings, or search for Everythings under Customize › Connectors in claude.ai or Claude Desktop.
- Click Connect.
- Sign in to Everythings on the page that opens. It lists the access Claude asked for (read, or read and write); press Approve.
- Start a chat and ask: “Show me my workspaces in Everythings.”
On a build without the directory, choose Add custom connector under Connectors and paste the server address instead:
https://www.everythings.app/api/mcp
Where it works
The connector belongs to your Claude account, so one connection serves every Claude surface that uses connectors.
- claude.ai and Claude Desktop: the tools and the inline view.
- Cowork: the tools, so a longer task can file its results into your workspaces as it goes.
- Claude Code: the Everythings plugin adds the same server to every project with its own sign-in, plus the
/thingspane and the band above the prompt. - Claude Desktop's config file starts local servers only. A remote server like this one goes in through Connectors, so there is no snippet to paste there.
What you see
Ask Claude to show a workspace or a thing and the answer arrives as the Everythings view, drawn in the conversation: a workspace as a grid of emoji tiles, a thing with its Markdown, its sub-things, marks and comments.
It is yours to click
Click a tile to open that thing. The breadcrumb across the top walks you back up, the workspace picker switches between workspaces, and open in app ↗ hands the thing to the full app. Every click fetches fresh data through your connection.
You see what Claude wrote
After creating or editing a thing for you, Claude shows it back in this view, in place in your workspace, so you can approve it or ask for a change while the context is still on screen.
Claude reads quietly
When Claude needs a thing's content for its own work, it reads it with plain tools and shows nothing. The view appears when you ask to see something, or right after a write you asked for.
Full detail, including the one thing the view can write (a checklist tick) and why it shows an agent's question read-only: The view in the chat.
Things to say
“Show me my workspaces in Everythings”
Claude draws your workspaces and your default workspace's tiles in the chat. Click a tile to open that thing.
“Plan a trip to Spain in my Everythings”
Claude creates a Spain thing with nested Tickets, Hotels and Itinerary things, fills each with its research, and shows you the result in place.
“Add milk and eggs to my groceries thing”
Claude finds your Groceries thing by name, reads its current list, and saves it back with the two items added.
“What did we decide about the roof?”
Claude searches your things for the topic, reads the right one, and answers from it without you opening the app.
“Ask me in Everythings which design I prefer, then carry on next time”
Claude attaches a question with options to the thing and stops. You answer on your phone; the next run picks the answer up.
People name a thing after its topic and refer to it that way, so “my groceries thing” and “the Spain thing” are how Claude expects to hear about one. Saying in Everythings once in a chat is enough to point it at the connector.
Approvals and access
Claude asks before using a tool according to its own permission settings. What the tool may do is set on the Everythings side.
- Read access lets Claude search and read; nothing changes. Read and write adds creating, editing, moving, copying, deleting to Trash and restoring things, comments, marks and questions. Claude asks for what it needs when it connects: if you are signed out, a page names Claude and asks you to sign in; then the approval page lists that access before you approve.
- Everything Claude writes carries its label: ✦ Claude on the thing and ✦ on its tile, “via Claude” on a comment, and “agent” after your name on a mark.
- A delete goes to Trash. Claude cannot empty Trash, delete a workspace or invite anyone.
- To disconnect, revoke its row under Settings › MCP Access Tokens in Everythings; removing the connector in Claude afterwards tidies Claude's list. See Connect an agent for the full list of what an agent can and cannot do.
Claude Code
In Claude Code, connect with the Everythings plugin. It signs in with OAuth, puts the tools in every project, and adds the /things pane, the band above the prompt and a session that carries on when you answer its question.
Install steps, the pane, its permission rules and troubleshooting: Claude Code plugin.
Headless Claude Code (servers, CI): an access token
For a run where nobody is there to sign in, such as claude -p on a server or in CI, give Claude Code an access token. Replace nthings_mcp_YOUR_TOKEN with a token from Settings › MCP Access Tokens:
claude mcp add --transport http everythings https://www.everythings.app/api/mcp \ --header "Authorization: Bearer nthings_mcp_YOUR_TOKEN"
Skip this where the plugin is installed: it already adds a server named everythings, and a second one with that name would clash with it. More on tokens: Access tokens & CLIs.
Questions
Why does Claude draw my workspace as tiles?
Everythings ships an MCP App. The two display tools carry an interactive view that Claude renders inline, so you get tiles, Markdown, marks and comments in place of JSON. It reads your data over the connection you already authorized. The view in the chat has the details.
Can the view change my data?
Only by ticking a checklist box, which saves the thing over the agent's connection and needs read & write access. Clicking tiles, breadcrumbs and the workspace picker only reads. A question an agent asked shows in the view read-only; you answer it in the app or by telling Claude.
Does it work in Claude Cowork?
Yes. A Cowork task uses the same connector, so it can research and file everything into your workspaces. To watch it build, keep the /everythings:panel page from the Claude Code plugin open beside it.
Claude says the connector is not connected.
Open Customize › Connectors, find Everythings and connect it again; a revoked or expired sign-in shows this. In Claude Code, run /mcp and sign in to everythings.
How do I disconnect?
Revoke its row under Settings › MCP Access Tokens in Everythings. That ends every token the connector holds for you, refresh tokens included, so Claude is disconnected within a minute. Removing the connector under Customize › Connectors in Claude afterwards only tidies that list.