See Scope Architect on your own project. Get a walkthrough →
Cloud Runs

Your agent builds. You review the PR.

Dispatch a task from your scope and get a pull request back. Each run gets a fresh, isolated VM with your agent briefed on the live plan — your project's workspace persists between runs, and nothing sits idle while you're not building.

Get StartedSee pricing
How it works

Three steps from task to pull request.

No local setup. No running agents overnight on your own machine. Pick a task, fire it off, and come back to a PR.

Step 1

Connect your GitHub account.

Link a repo and authorize Scope Architect to open pull requests on your behalf. Builder plans connect one account. Team plans connect multiple.

Step 2

Dispatch a task from your scope.

Pick any task from your project plan. The agent receives the full context: your plan, tech stack, prior decisions, and acceptance criteria.

Step 3

Review the pull request.

The agent builds inside a fresh, isolated VM, commits the work, and opens a PR. You review it, request changes, or merge when you're ready.

The SA difference

Your agent, but already briefed on your entire project.

Most cloud coding tools start from the repo and a prompt. Cloud Runs start from your live scope, so your agent knows what you're building, why, and what matters — before it reads a line of code.

  • Reads your full project plan and module structure before it starts
  • Knows your stack, preferences, and prior decisions from your Builder Profile
  • Works against the task's acceptance criteria, not a vague prompt
  • Posts updates and questions back to the project as comments
Context at startOther cloud tools: repo + task onlyCloud Runs: full plan, decisions, stack
Task definitionOther cloud tools: freeform promptCloud Runs: scoped task with acceptance criteria
Visibility while it buildsOther cloud tools: black box until doneCloud Runs: live comments and status in your workspace
How runs work

Fresh VM every run. Workspace that persists.

Every run spins up a clean, isolated VM — no leftover state, no idle servers burning while you sleep. Your project's workspace lives on a saved disk, so the next run picks up right where the last one left off, instantly.

Agent flexibility

Your agent. Your account. No markup.

Cloud Runs use the agent CLIs you already use — connect Claude Code or Grok, or add your team's own API keys. Runs bill to your account either way, with no markup and no metering from us. Codex, Antigravity, and Cursor Cloud are on the way.

Claude CodeGrokTeam API keysCodex — coming soon

No markup. No metering. You own the provider relationship.

Plans

Included on all paid plans.

Every paid plan includes concurrent cloud runs — one task building at a time per slot, across any of your projects. Need more running in parallel? Add concurrent runs at +$19/mo each.

Builder · $49/mo

Solo builders and indie developers.

  • 1 concurrent cloud run
  • 1 GitHub account
  • Claude Code, Grok, or team API keys
  • Full MCP integration
  • +$19/mo per extra concurrent run
Team · $79/seat/mo

Teams building with AI across multiple projects.

  • 1 concurrent cloud run per seat
  • Multiple GitHub accounts
  • Claude Code, Grok, or team API keys
  • Real-time collaboration and role permissions
  • +$19/mo per extra concurrent run
Get Started

Dispatch a task. Come back to a PR.

Cloud Runs execute inside your scope, with full context, and push to GitHub when done. You stay in control.

Get StartedSee Agent Handoff