Ads Acquisition Engine
From brief and website to a full paid-acquisition plan, offline.
An offline-first TypeScript engine that turns a brief and a website into a paid-acquisition plan — strategy, tracking, creative with QA, campaigns, experiments and diagnosis — through a CLI, an HTTP API and an MCP server.
The engine reads a brief and a landing page, infers the funnel, call to action and value proposition with flagged confidence, then produces channel strategy with feasibility warnings, a tracking plan and attribution model, creative concepts and platform-fitted variants checked by creative QA, campaign architecture and experiments with sample sizes. Imported performance data is diagnosed and turned into recommendations.
State lives in SQLite and every non-computed fact carries its source. Sending anything to a platform needs an approval bound to a hash, credentials and a registered transport; version 1 has none, so `npm run demo` runs the whole pipeline offline and contacts nothing.
The demo runs on LumenDesk, a fictional brief included in the repository, with synthetic performance data.
- Type: Paid-acquisition planning engine (CLI, API, MCP)
- Status: In development — Version 1 with a test suite. It plans and exports; it ships no live connection to ad platforms, and execution is a dry run by default.
- Code: simonvildonelohim/ads-acquisition-engine (private repository)
- Year: 2026
Built
- Website analysis: offer, CTA, funnel, proof, friction
- Channel strategy with feasibility warnings
- Tracking plan and attribution model
- Creative concepts and variants with QA
- Campaign architecture
- Experiment planning with sample sizes
- Performance diagnosis and recommendations
- Hash-bound approvals; dry run by default
- CLI, HTTP API and MCP server
Technology
TypeScript, Node.js, zod, SQLite, MCP, Vitest
