Mission control for building with AI.
A detailed scope in minutes. Every agent's progress on one board. Works where you already work. No context lost from chat to chat.
Compatible with
+ more
Idea becomes an agent-ready . Manage and track the as agents work. Agents pick it up in your .
Scope it
A detailed scope, in minutes. Chat it through or paste a brief. Modules, tasks, acceptance criteria, dependencies — a blueprint an agent can actually execute. Then keep shaping it in Architect Chat.
Scope it →Track it
Manage and track progress as your agents work. Every module, task, and status on one board — updated as they go. Switch chats, switch agents: nothing to re-explain.
Track it →Build from it
Drop it into the workflow you already have. Point Claude Code, Cursor, or Codex at the plan and they work it from your editor. Or queue a task and a cloud agent comes back with a pull request. New chat, same project.
Build from it →Everything flows through the scope.
Not a markdown file in your repo. Every agent reads the scope, writes back what it did, and you see it without opening the chat.
Think it through. Then hand it off.
Talk the idea out with Architect Chat like a whiteboard: pressure-test it, cut the MVP, see the real surface area before anything touches code. Then import the repo and the Scope Engine turns what's left into modules, tasks, and acceptance criteria that match your stack. The plan lives here, not in a chat log.
Scope it →- Architect Chat
- Scope Engine
- Grounded in your repo
- Acceptance criteria on every task
- Defaults that travel
One board every agent reads and writes.
Chat forgets everything when the session ends, and every agent starts from zero. Scope Architect is the layer they all read and write: the plan, the memory, and the live state. Switching from Claude Code to Cursor costs you nothing. And a teammate today, or you tomorrow, sees the same board without reading anyone's chat.
Track it →- Live statuses & comments
- Agent asks, you answer
- Memory that carries
- Who did what, human or agent
- What it cost
In your editor or the cloud. Ends in a PR.
Work the hard module yourself in Cursor, with the plan in hand. Queue the rest from your phone and wake up to a pull request. Every run is approval-gated, streams live, and stops at the budget you set. Unattended is not unsupervised.
Build from it →- Claude Code, Cursor, Codex
- Queue from chat or your phone
- Approval gates and budgets
- Live streaming logs
- PRs back to GitHub
Also in the workspace.
Architect Chat
Talk the idea through. Changes come back as plans you preview and approve.
Grounded in your repo
Import the codebase. Scopes land in your stack, not a greenfield fantasy.
Every editor, one plan
Claude Code, Cursor, Codex. Nothing to install. Switch agents and nothing is lost.
Cloud runs → PR
Queue a task. A cloud agent builds it and opens a pull request.
Memory that carries
Context, decisions, and docs live with the project. Next session starts caught up.
Team on the same board
Humans and agents, including people who never open an editor. One plan.
“Scope Architect is the home base for my whole build. Your plan shouldn't live only in the code.”
The latest
- OCT 10, 2026Projects and scopes: one name for each level
What the app called a Product is now a Project, and what it called a Project is now a Scope. Two levels, one name each, the same everywhere: Project › Scope › Modules › Features › Tasks. Nothing about your data changed.
- OCT 10, 2026Pricing: Builder $49, Pro $99, Team $49 a seat
Builder is $49 a month, Pro is $99 with three concurrent cloud runs, and Team is $49 a seat with a two seat minimum. Annual billing is still two months free.
- SEP 02, 2026Launch pricing: Builder $30, Team $50 per seat
Builder is now $30 a month and Team is $50 per seat, with annual billing at two months free. Every plan includes unlimited products, hosted MCP, cloud runs with budgets, and your own API keys with no markup. Existing subscribers keep the price they signed up at.
Youyour agentsandanyone who needs to change the product. Same board. PRs back.