Documentation
Sprawl runs your operations for you — each one walled off on its own accounts, with its own fleet of agents. Start from the coding agent you already have open, or from the browser. Everything here describes how Sprawl actually behaves today — no aspirational docs.
Your first five minutes — from your coding agent
One command installs Sprawl into Claude Code, Codex, or any MCP client; your agent authorizes in your browser over OAuth, then designs and stands up an operation from a sentence. The measured walk, including the one step it cannot do for you.
A weekly brief before you have customers — nothing to connect
One recurring job, set up from your coding agent in one sitting: who might pay, and what one experiment tests it, every Monday. The exact blueprint, the tool calls in order, and what came back when we walked it on production — including what the brief cannot read.
Getting started in the browser
No coding agent? The same thing from the web app: start from what you’ve built (or describe it), approve one blueprint, activate the fleet, and put it on autopilot.
Connecting your accounts
Give your agents access to your tools — scoped per operation (a different Gmail, Slack, or X each), granted through each service’s own sign-in, and reversible anytime. 100+ integrations.
How automations run
The ways your agents run work on their own — on a schedule, on an event trigger, as a goal-directed pursuit that paces itself, or on demand. All in the background, all under the same approvals.
How approvals work
Agents act on your real accounts — approvals keep you in control. Routine work runs on its own; anything reaching the outside world waits in your inbox with a preview you can edit before you approve; the riskiest actions never run autonomously.
Questions the docs don't answer yet? The fastest way to get help is to start a free account and ask Sprawl directly in the app. Also see the FAQ.