A vague idea in.
An agent-ready plan out.
Drop in an idea, a ticket, or a repo. The Scope Engine decomposes it into modules, features, and atomic tasks with acceptance criteria, each sized for a single agent session and calibrated to your stack.
scope engineFrom a raw idea to an agent-ready plan.
Dump your idea, ground it in your real code, let the engine assemble the modules, and finalize the pre-scope, all before a single task is generated.
Dump your idea
Clear on what to build? Dump your messy notes or a full project brief. Not clear? Shape it through guided discovery with Architect Chat.
Import an existing project
Connect a Git repo so your scope is grounded in your real stack, not a blank slate.
AI builds your modules
The engine reviews your idea, then auto-selects and assembles the modules it should contain.
Configure and finalize your pre-scope
Review the pre-scope, adjust modules and settings, and lock it in before a single task is generated.
01.Drop in your idea
Paste raw notes, a Linear ticket, a Slack thread, or just a one-liner. The engine handles parsing, context extraction, and initial requirement mapping — no rigid templates required.
02.Pick your modules
Pre-configured starting points (auth, billing, dashboards, APIs) ground the engine in reality. Your Builder Profile (stack, target agent, code style) shapes every spec it produces.
03.Recursive decomposition
Multi-pass recursive expansion maps vague requirements into modules, features, and atomic, agent-sized tasks with acceptance criteria and dependency graphs.
04.Refine in Architect Chat
Edit modules and tasks inline, or talk to the architect — "break this finer," "add edge cases," "mark agent-ready." Every change is a diffable plan you preview and apply in one click.
A complete scope, ready to build.
Modules, features, and atomic tasks, generated, refined, and shaped together with your team.
Your AI product partner, in every build.
Talk through scope against time and risk. Float an idea before you commit and it maps the real surface area. Ask what to cut and it reasons over the whole plan, not just the task in front of you. Every change comes back as a structured plan you preview and apply. No drift, no half-applied edits.
- Decide what's in the MVP
- Pressure-test a new feature
- Reprioritize when scope creeps
- Split work into agent-ready tasks
- Approval-gated scope edits
- Queue tasks and launch runs from chat
Your defaults, built into every scope.
Next.js or Remix, Postgres or Mongo, tRPC or REST: pin your defaults once and every scope lands in a context your agent already understands. Name your target agent and the engine tunes acceptance criteria and task granularity to match. Hours, story points, or agent sessions: the math stays consistent, the labels match how you plan.
- Tech stack
- Target agent
- Estimate unit
- Code style notes
- Testing preferences
- Acceptance-criteria style
Everything behind your build, in one place.
Builder Profile
Your stack, target agent, code style, and estimate unit. The engine scopes to your defaults on every project.
Connect GitHub
Link repos to projects so scopes are grounded in your real code and agents can open PRs against it.
Preset Library
Reusable module and feature presets that drop straight into a scope, so common builds start most of the way done.
Export anywhere
Export any scope to clean JSON or Markdown, ready to hand to any tool or agent.
Context Dashboard
Always-injected context blocks your AI reads on every generation and build, so nothing gets lost.
Memory Management
Durable facts your agents record and reuse, so the project keeps its memory between sessions.
Markdown Docs
Markdown docs that live with the project. Pin the ones your agent should always read.
BYOK
Bring your own keys for Anthropic, OpenAI, Google, xAI, or DeepSeek. Pay providers directly, with no markup.
Team Management
Invite your team, manage seats and roles, and work the same plan together in real time.
Connect CLI, run in the cloud
Connect Claude Code or Grok to run tasks in the cloud and get a pull request back — more agents coming.
Token Analytics
Per-operation token usage and cost across every project, in one usage dashboard.
Take command of your next build.
Your agents do the building. You stay in control of every move.
Bring your API keys.
Run on platform credits out of the box, or bring your own API keys for Anthropic, OpenAI, Google, and more to go unlimited.