Inspiration
Having a startup idea leaves you with a lot of work to figure out.
What should the product do? Who is it for? What goes on the website? How do you explain it to someone who has never heard of it?
Campaign AI brings those early steps into one workspace. You start by describing your idea, then work through the plan, website, marketing content, and product pitch.
What it does
The starting point is a plain-language description of what you want to build or promote.
Campaign AI extracts the goal, topic, and target audience. It uses web research to develop an audience profile and core messaging, then prepares a strategy broken into phases.
It also generates a business requirements document covering the product overview, objectives, proposed features, user stories, and success metrics. The document can be downloaded as a PDF.
From there, the workspace has three other parts:
Website: Generates a promotional landing page using the audience and messaging context. The web editor lets you change the HTML, see a live preview, copy the code, and request deployment to Vercel when deployment credentials are configured.
Social content: Creates social-post drafts with supporting stock images and displays them in Instagram- and X-style previews.
Pitch Studio: Takes a product name and website link, reads the website, and prepares a prompt for a conversational sales assistant. The interface also includes a Vapi voice-demo flow with transcript and call-event displays. The generated prompt is copied into the Vapi assistant settings manually.
For example, someone working on an inventory tool for small shops could use the workspace to outline the product requirements, prepare its first landing page, draft promotional posts, and prepare a product pitch.
How it is built
The frontend uses React, Vite, and Tailwind CSS. Python and FastAPI handle the backend.
LangGraph connects the planning, research, strategy, content, design, website, document, and operations stages. These stages share a campaign state, so the audience and messaging can carry through to later outputs.
Groq serves Llama 3.1 8B Instant, Tavily provides web search, Unsplash supplies stock images, and fpdf2 creates the requirements document. WebSocket updates let the frontend display results as the workflow progresses.
A separate service reads product websites and generates the voice-assistant prompt.
Current prototype
The repo contains the generation workflow, research view, strategy breakdown, document export, website editor, social previews, and Pitch Studio interface.
Some integrations are still partial. Direct Instagram and X scheduling is simulated, and the meeting-booking service currently returns a mock Calendly response. Slack and Telegram delivery code is available when the relevant credentials are configured.
The generated website is a promotional landing page. The current website prompt excludes forms and interactive signup controls.
Campaign AI’s broader aim is to make the first steps of launching a startup easier to work through: understand the idea, put it into a plan, build a page around it, and prepare how to introduce it.
Built With
- blake3
- css
- html
- httpx
- javascript
- python
- sqlite
- typer
- zstandard
