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

Pricing built for builders.

Start free, upgrade when the work gets real. Flat pricing, unlimited products on paid tiers, and BYOK — pay providers directly, no markup.

Free

$0
free forever
  • 1 ProductOne product with its full scope: modules, features, and tasks across as many projects as it needs. Upgrade for unlimited products.
  • Full MCP read/writeYour agent connects to the live workspace, pulls ready tasks, and writes status, comments, and memory back as it builds. The full loop, not a demo.
  • 150 credits / moEnough to scope a product and refine it. 1 generation = 50 credits, chat = 1, AI apply/edit = 5.
  • Agent Handoff exports
  • 1 GitHub account
  • Cloud runs
  • BYOK

Scope one product end to end and connect your agent over MCP. No card required.

Builder

$49/mo
  • MCP connection
  • 1 concurrent cloud runDispatch a task and get a PR back. Each run gets a fresh, isolated VM; your project's workspace persists on a saved disk, so the next run resumes instantly. No idle servers. Add extra concurrent runs for +$19/mo each.
  • 1,000 credits / moTypical month: 2-3 generations + heavy chat = ~200 credits. 1 generation = 50 credits, chat = 1, AI apply/edit = 5. BYOK removes the meter entirely.
  • Premium AI modelsRun generations and chat on premium models (Claude, GPT) at higher credit rates, or use included models at standard rates.
  • Unlimited products & projects
  • Agent Handoff exports
  • 1 GitHub account
  • BYOK - no credit limitsConnect your own API keys and pay providers directly at their published rate. No markup, no metering. Cloud runs stay subject to fair-use run-time limits.

MCP connection, cloud runs, BYOK, and the full toolkit for solo devs shipping with AI agents.

Team

$237/mo
$79/seat × 3 seats
Team seats3
  • Everything in Builder
  • 1 concurrent cloud run per seatRuns scale with your team: 3 seats = 3 runs at once, one per project at a time. Add extra concurrent runs for +$19/mo each.
  • 1,000 pooled credits/seat/moCredits pool across the team. 1 generation = 50 credits, chat = 1, AI apply/edit = 5. BYOK removes the meter entirely.
  • Multiple GitHub accounts
  • Real-time collaboration on scopes
  • Role permissions

Everything in Builder, shared across your team. Real-time collab, pooled credits, and role permissions.

Bigger team or heavier usage? Talk to us about custom billing.

Included in every plan.

The core workflow ships at every tier.

Scope Engine

Turn a vague idea into modules, features, and atomic tasks your AI can actually execute.

Architect Chat

Refine and direct your scope by talking to AI. Diffable change plans you preview and apply in one click.

Command Center

One real-time view of every task: status, owners, agents, and time in progress across the whole project.

Project Memory

The context, memory, and docs your AI reads and writes as it builds, so nothing gets lost between sessions.

Builder Profile

Your stack, your target agent, your code style, your estimate unit. The engine calibrates to your defaults.

MCP connection

Live link to Claude Code and Cursor. Your AI reads the current plan directly and writes back, no copy-paste required.

Agent Handoff exports

Clean JSON, Markdown, or rules optimized for Cursor, Claude Code, Devin, or Windsurf. Copy, paste, ship.

Not sure yet?Get a walkthrough →See it scoped on your own project, live.
Why Scope Architect

Why not just write a long prompt in Cursor?

You can. And it'll work for a small feature. Here's what changes when the project gets real.

Raw prompting
  • Re-explain the project every session
  • Agent invents its own task boundaries (badly)
  • Acceptance criteria are vibes, not contracts
  • No shared plan between humans + agents
  • Edge cases discovered mid-implementation
Scope Architect
  • Project context lives in the plan, not the prompt
  • Recursive decomposition gives atomic, agent-sized tasks
  • Explicit acceptance criteria on every task
  • Dependency graph + shared Builder Profile
  • Edge cases surfaced before the agent starts spending tokens

FAQ

Scope Architect is mission control for building with AI. Scope a project into agent-ready tasks, hand it to your coding agent over MCP or a cloud run, and stay in command as the build moves: context stays, progress is visible, and every decision is surfaced in one workspace.
Any of them. Connect Claude Code, Cursor, Windsurf, Codex, and more over MCP so the agent reads and updates the live plan directly. Or dispatch a task to a cloud run and get a pull request back. For everything else, export a clean JSON or Markdown spec tuned to your target agent.
Three paths. MCP (recommended): your agent connects to the live workspace, pulls the next ready task, and writes status, comments, and memory back as it builds. Cloud run: dispatch a task to a fresh, isolated VM and get a pull request back — your project's workspace persists between runs. Manual export: copy a self-contained JSON or Markdown spec into any agent for a one-off handoff.
Yes. Connect your Anthropic, OpenAI, Google, xAI, or DeepSeek keys and pay providers directly at their published rates. No markup, no credit meter. Your keys are encrypted at rest and your project context stays under your control.
Yes, two ways. Start free: scope one product end to end with full MCP, no card required. Or get a walkthrough and we'll scope a real project together and kick off a cloud run live, so you see exactly what you get.
A reusable set of preferences: your tech stack, preferred agent (Cursor / Claude Code / Devin / Windsurf), code style notes, estimate unit (hours, story points, agent sessions), and testing preferences. The engine reads it on every generation so specs land in a context your agent already understands.
Yes. Tasks have statuses (To Do, In Progress, In Review, Done, Blocked) and live on a shared plan. Humans pick tasks; agents pick tasks; both work in parallel against the same spec. Agent handoffs export per-task so you can scope just what you want shipped.