AI PRD Generator

Free tool vs. full Tekk

The free tool gives you the structured first draft. Tekk grounds the rest of the work in your actual codebase.

How it works

1. 1

   Describe the product

   Give the tool a one-line name, the problem you're solving, who it's for, and any constraints worth knowing.

2. 2

   Generate the PRD

   We send your inputs to a tuned PRD prompt. The document streams back as editable markdown so you can read it as it writes.

3. 3

   Continue in Tekk for the real spec

   One click hands the PRD into Tekk, where you connect your repo and turn it into an executable plan with subtasks, acceptance criteria, and Kanban tracking.

Frequently asked questions

What is a PRD?

A Product Requirements Document is a written description of what you're building, who it's for, why, and how success will be measured. PMs use it to align engineering, design, and stakeholders before any code gets written. Done well, it answers every question someone could have about scope before they ask.

What sections does a good PRD include?

The shape varies by org, but a strong PRD usually covers: a TL;DR, the problem, goals and non-goals, target users, requirements (often grouped by Must/Should/Could priority), success metrics, risks and assumptions, and an explicit out-of-scope section. The Non-goals and Out-of-scope sections are the most underrated — they're how you prevent scope creep before it starts.

PRD vs MRD vs BRD — what's the difference?

A BRD (Business Requirements Document) is the strategic 'why we're doing this' — usually owned by execs or sales. An MRD (Market Requirements Document) is the 'what problem exists and who has it' — usually owned by product marketing. A PRD is the product-level 'what we'll build to solve that problem' — owned by the PM. At smaller companies the PM typically just writes the PRD and skips the other two.

How long should a PRD be?

Short enough to read in one sitting, long enough to leave no major ambiguity. For most features that's 1–3 pages. If it's longer, the feature probably needs to be split into multiple PRDs.

Who writes a PRD?

The product manager, usually. At smaller startups, the founder or lead engineer writes it. The point is one accountable author — not a doc written by committee. Other people review and contribute, but one person owns the source.

Can I edit the result?

Yes — copy the markdown out and edit it anywhere. PRDs are living documents; this tool gives you a structured first draft, not the final answer. For an editable document with version history, follow-up questions, and codebase context, hit 'Continue in Tekk' to take the PRD into Tekk's planner.

How is this different from ChatGPT?

The prompt is tuned for PRD conventions (problem → users → goals → non-goals → requirements → metrics → risks). You get a consistent shape every time. ChatGPT can produce a PRD too, but you'd have to remember to ask for each section. Tekk's full planner goes much deeper than either — it reads your actual codebase and structures the PRD around your real files, models, and patterns.

How is this different from Tekk's full planner?

This free tool generates a static PRD from your text description. Tekk's planner reads your actual codebase (GitHub, GitLab, or Bitbucket), references your real files and patterns in the PRD, asks you follow-up questions, splits the PRD into executable subtasks with acceptance criteria, and tracks the work on a Kanban board. The free tool is the first 10% of what Tekk does.

Does Tekk store my PRD inputs?

The free tool doesn't store anything. Your inputs go to OpenAI for generation, then the result is rendered in your browser. If you click 'Continue in Tekk', the markdown is staged for 10 minutes so you can claim it after signup — after that it's deleted. Plans you create after signing up are stored in your Tekk account, same as any SaaS.

Why only two free runs?

AI generation costs us money on each request. Two runs is enough to convince yourself the output is useful before signing up. Signing up takes 30 seconds, connects your GitHub for codebase-aware planning, and unlocks unlimited generation.

What model do you use?

OpenAI's gpt-5-nano, with the minimum reasoning effort for speed. It's the cheapest model in OpenAI's current lineup that produces structured, decisive output. The system prompt does most of the lifting — we tell the model the exact section order, voice, and conventions.