Skip to main content
The dashboard generates the right command for your tool and watches for the connection, so the real setup happens there. Here’s the shape of it, for every tool.

​
The flow, for every tool

1

Pick your tool

Sign in at app.atlaso.ai and pick your tool (or open Connections if you’re already set up).
2

Run the command

Run the command it shows you — or paste the connector URL for GUI apps like Claude Desktop.
3

Authorize in the browser

A browser window opens by itself — that’s the device authorization step. Check it’s app.atlaso.ai and click Authorize.
4

Done

The tool starts capturing as you work and recalling before your prompts — nothing else to configure.
The browser opening on its own the first time a tool runs is expected — it’s how the tool links this device to your account. It only ever opens your Atlaso authorization page.

​
What each tool supports

CLI and IDE tools get the full automatic loop. GUI chat apps connect through MCP, which works differently — the model calls memory tools when it decides to (see Claude.ai & GUI apps).
ToolAuto recallAuto captureAmbient (Pro)Notes
Claude Code✓✓✓Full support: hooks + MCP + skill
Codex CLI / Desktop✓✓✓Desktop installs via its marketplace UI
Cursor✓✓✓Recall arrives via a generated rules file
OpenCode✓✓✓Pure plugin, npm-distributed
Antigravity✓✓✓Restart after install so plugins re-scan
Claude Desktop———MCP tools only — model-initiated memory

​
On the Free plan

Free includes one device and one active tool — the first tool you use claims the slot automatically. A second tool keeps working locally but stops syncing to your cloud memory. How that works and how to switch: One active tool on Free.