Connecting your accounts
Your agents do real work in your real tools — your inbox, your Slack, your repos. Connections are how you grant that access: scoped to one operation, granted through each service’s own sign-in, and reversible whenever you want.
One operation, its own accounts
Every operation is a sealed container with its own connected accounts. You can connect a different Gmail, Slack, or X to each one. Run a company and a side project, or five separate clients, and nothing bleeds across — the account one operation’s agents use is never the account another’s do. This is the whole point of the per-operation model: no shared blob of credentials, no crossed wires.
Connect from the checklist
After you approve a blueprint, a connect checklist shows exactly what each agent needs. Link an account and it flips to ✓, bringing its agents online — one sign-in often activates several agents at once. You can connect before or after provisioning; agents wait patiently for the access they need.
It’s your sign-in — Sprawl never sees your password
Every connection goes through that service’s own OAuth consent screen — Google, Slack, GitHub, and the rest. You grant exactly the access the screen shows, and Sprawl holds a revocable token, never your password. If a provider offers granular scopes, you choose them on their screen, not ours.
Skip anything you’re not ready for
Not ready to connect something? Skip it. Its agents stay parked until you do — a half-configured agent never fires a broken run. Automations that depend on a disconnected tool are skipped for that run and noted, rather than failing loudly, and pick back up the moment you reconnect.
Connect the same app more than once
Because connections are per-operation, you can link a different Gmail — or Slack, or X — to each operation without them colliding. The account your client-work agents use isn’t the one your personal agents use. Connect an app as many times as you have operations that need it.
Disconnect anytime
Revoke a connection whenever you want. The agents that relied on it simply park until you reconnect — nothing an agent does keeps running without live, granted access. Control stays with you, continuously, not just at setup.
A hundred-plus integrations
Gmail, Slack, GitHub, Linear, Notion, Google Calendar and Sheets, and dozens more are available out of the box. Under the hood Sprawl uses Composio and MCP for the integration layer, so the catalog keeps growing. Building your own agent elsewhere? It can connect to Sprawl over the MCP server and use the same tools.
Next steps
New here? Start with Getting started to see where connecting accounts fits in the flow, and read how approvals work to see how much control you keep once your agents are live.