Agency Engine turns a client transcript into a Claude Code-ready build package.
Drop in transcripts, SOWs, or raw kickoff notes. Get back a sitemap, section content, schema, and build phases — the complete package Claude Code needs to start without guesswork.
$ agency-engine --run intake ✓ Parsing intake documents... ✓ Processing brand assets (3 files)... ✓ Extracting requirements... ✓ Generating sitemap... pages/ ├── home.md ├── about.md ├── services/ │ ├── web-design.md │ └── development.md └── contact.md ✓ Generating content... ✓ Building plan package... ✓ Scaffold ready. Output: ./agency-engine-output/ Documents: 8 files generated Time: 47s
Intake goes in. Build package comes out.
One pipeline. Every document your AI code agent needs to start building.
Drop in your raw client intake — transcript, scope of work, notes, whatever you have. Attach brand assets with a context note — logo files, a guidelines PDF, inspiration screenshots — and they're passed as vision inputs that inform design system generation. Agency Engine structures it all, reasons over it, and produces a complete plan package. Sitemap, page-level content, component specs, schema definitions, build phases. Not a summary. Not a prompt. A real set of technical documents ready to hand directly to Claude Code.
# Sitemap — Acme Roofing Co.
pages:
- id: home
url: /
sections: [hero, services, testimonials, cta]
- id: services
url: /services
sections: [header, service-grid, process, cta]
- id: contact
url: /contact
sections: [header, form, map]Five stages. One build-ready package.
A structured pipeline that takes raw client input and produces everything you need to start building — in the right order, every time.
Intake
Drop in your transcript, SOW, notes, and brand assets.
Sitemap
AI generates a structured sitemap. You review and approve.
Content
Section-level copy generated for every page.
Plan Package
Eight technical build documents, ready for Claude Code.
Scaffold
Provision services, push to GitHub, and hand off to Claude Code.
Not a wrapper. Not a prompt tool. A pipeline.
STRUCTURED OUTPUT
Every stage produces a real, structured artifact — not a prompt you have to clean up or feed somewhere else. The output is the deliverable.
schema.ts ← Data models build-phases.md ← Build sequence component-map.md ← UI components
CLAUDE CODE–READY
The Plan Package is built specifically for Claude Code. Drop it in and start building — no reformatting, no translation, no prompt engineering required. Provisioned service credentials are injected into the starter prompt automatically.
agency-engine-output/ ├── sitemap.md ├── content.md ├── schema.ts └── build-phases.md
PRACTITIONER-BUILT
Agency Engine wasn't designed in a vacuum. It was built to solve the actual planning overhead that slows down real client projects — by someone running those projects.
Built for practitioners doing real client work.
Agency Engine is not for agencies with dedicated PMs and 12-week timelines. It's for the one-to-three person shop where the person doing the design is also writing the spec and setting up the repo.
It's for the freelancer doing five to ten sites a year who wants to move faster without cutting corners on planning. If your current process involves a shared Google Doc, a voice memo from a client call, and a lot of tab-switching — this is what replaces the middle part.
This is for you if…
- →You run a small agency or solo freelance practice
- →You do client work — real projects, real scope documents
- →You're already using or interested in AI-assisted development
- →You spend too much time on planning docs before a line of code is written
- →You want consistent structure across every project you take on
- →You're building with Next.js or a similar modern stack
Common questions about data and API keys.
Is my API key safe?
Your Anthropic API key is encrypted at rest using AES-256-GCM encryption before being stored. It is never stored in plaintext, never visible to us internally, and never returned to your browser. You can remove it from your account settings at any time.
Who pays for the AI generation?
You do, directly. Agency Engine uses your own Anthropic API key — we don't provide AI compute and we don't mark up API costs. A typical full pipeline run costs approximately $0.50–$2.00 on Claude Sonnet.
Do you train on my data?
No. Your project content is processed by Anthropic's API using your own key and is not used to train any models. We do not use your inputs or outputs for any purpose beyond running the pipeline.
What happens to my data if I cancel?
Deleting your account permanently removes all your data from our systems — projects, generated files, brand assets, and stored credentials. Deletion is immediate and irreversible.
Beta users are building with it now.
Agency Engine is currently in private beta. Practitioner feedback is coming — and it'll be real when it does.
Get in before public launch.
Agency Engine is opening to a limited group before the public beta. 14 days free, no credit card. First 50 founding members lock in at $9/mo forever.