Skip to main content
Free AI Development Planning Tool

Vibe Workflow

turns ideas into build-ready AI docs

Structure with AI uses Gemma 4 (Free AI) · @cf/google/gemma-4-26b-a4b-it and counts toward your provider or plan quota.

Continue directly, or explicitly use AI to extract target users and features.

Context compounds

AI needs a chain of decisions—not one heroic prompt.

Vibe Workflow carries evidence forward. Research shapes the PRD, the PRD constrains the architecture, and the architecture becomes an executable build plan your coding agent can actually follow.

What the website does

It interviews you at your technical level, generates each artifact with your chosen model, lets you review and refine the result, and exports the full project context without locking it inside the app.

  1. 01

    Research the idea

    Validate the problem, map competitors, surface technical constraints, and preserve sources for later decisions.

    RESEARCH.md

  2. 02

    Define the product

    Turn evidence into users, scope, requirements, acceptance criteria, and a launch definition the build can follow.

    PRD.md

  3. 03

    Design the system

    Choose the stack, data model, APIs, security boundaries, deployment shape, and implementation constraints.

    TECH_DESIGN.md

  4. 04

    Plan the build

    Sequence the work file by file with phases, dependencies, verification commands, and completion criteria.

    BUILD_PLAN.md

  5. 05

    Package the handoff

    Generate agent instructions and export the complete context package for Cursor, Claude Code, Codex, or your team.

    AGENTS.md + ZIP

Planning economics

One complete plan. Different model bills.

Estimated API spend from first research through PRD, architecture, build plan, agent instructions, and ZIP export. Your key, your provider bill, no markup from us.

Calls
5
Input
18k
Output
32k+

Thinking tokens are estimated from the selected reasoning profile. Actual bills vary with your idea, artifacts, retries, caching, and provider region.

Token estimate Search allowance Included quota
How this estimate works

Research

The typical case allows three grounded searches where Vibe Workflow currently wires provider search. Google’s first 5,000 shared monthly queries may be free.

Tokens

Typical means 18k input and 32k visible output tokens across five calls. Reasoning and tokenizer multipliers are applied per model.

Export

ZIP assembly happens locally and adds no model charge. Refinements, retries, regional premiums, and taxes are excluded.

Prices checked July 11, 2026. Qwen is converted from international CNY pricing at ¥7.20/USD. Claude’s published newer-tokenizer uplift is included. Cloudflare rows show $0 user out-of-pocket because this app supplies those models through plan quotas; underlying Workers AI list rates are not zero.

Before you start

Clear inputs. Portable outputs.

The workflow stays useful whether you are building your first app or handing a specification to an engineering team.

What is the vibe coding process?

The vibe coding process is a 5-step workflow: research the idea, write the PRD, plan the technical design, configure the coding agent, and then build with that context.

What is AGENTS.md used for?

AGENTS.md gives AI coding tools persistent project context, constraints, and implementation rules so they can build with fewer retries and less drift.

Who is Vibe Workflow for?

Vibe Workflow is for non-technical founders, working developers, and learners who want a structured AI-assisted path from idea to shipping.

How much does a complete plan cost?

The website is free to use. With BYOK models, you pay the provider directly for the tokens and optional search tools used across five generation steps. The homepage comparison shows transparent low, typical, and heavy estimates.

Start with the idea

Give your coding agent the context it deserves.

Move from a rough concept to research, requirements, architecture, an executable plan, and a portable agent handoff.