Docs

How SpecForge works

SpecForge turns raw customer signal into a structured, agent-ready PRD. Here's what happens between pasting your first interview and exporting a spec.

The pipeline

I

Ingest

Signal — interview notes, support tickets, PDFs — is chunked and embedded.

II

Cluster

Recurring themes across chunks are grouped into distinct opportunities.

III

Score

Each opportunity is ranked by how often it comes up and how severe it is.

IV

Review

You approve, edit, or reject opportunities before anything gets written.

V

Generate

Approved opportunities become a structured PRD with an agent-ready prompt.

Getting started

  1. 1.Create an account and start a new project.
  2. 2.Paste interview notes or support tickets, or upload a PDF, on the Signals tab.
  3. 3.Run the pipeline — clustering and scoring finish in under two minutes.
  4. 4.Review the ranked opportunities. Approve the ones worth building, reject the rest.
  5. 5.Generate the PRD and export it as markdown, or copy the agent-ready prompt straight into your coding agent.

FAQ

What counts as a "signal"?

Anything that carries real customer voice: interview transcripts, support tickets, survey responses, sales call notes. Paste text directly or upload a PDF.

Can I edit an opportunity before generating the PRD?

Yes. Opportunity titles and severity are editable inline on the review screen, right up until you approve and move to generation.

What does the exported PRD contain?

A summary, user stories with acceptance criteria, edge cases, an optional schema sketch, and a prompt written for a coding agent — exported as a single markdown file.

Is my data used to train anything?

No. Signal content is processed only to generate your PRD. See the Privacy page for details.

Start free trial