Atomics speaks MCP, the Model Context Protocol, so an assistant like Claude can read and edit your spaces as you. You add Atomics once as a connector, sign in, and choose what it may touch.
https://atomics.app/api/mcp
An assistant gets only what you approved, and acts as you: it reaches the spaces you can, and nothing else.
spaces:read | See your spaces and read the documents in them. |
|---|---|
spaces:write | Create and change your spaces and the documents in them. |
With read alone, the tools that change things refuse, and the shell runs read-only.
The same server answers in two shapes. Pick one when you add the connector; you can add both.
Standard/api/mcp |
One tool for each job: read a table, write a note, move a page on a board. The best fit for most assistants. |
|---|---|
Shell/api/mcp?profile=shell |
The terminal as one tool, plus the few things a line of text can’t do. Fewer tools to choose between and a smaller context, for agents at home in a command line. |
For the shell profile, add the connector with this URL instead:
https://atomics.app/api/mcp?profile=shell
Every tool still answers when it is called by name. The profile only changes which ones an agent is shown.
Next to the shell, an agent connected this way is shown the tools for:
Everything else is a line in the shell. See the shell’s commands