Vibe Workflow
turns ideas into build-ready AI docs
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.
- 01
Research the idea
Validate the problem, map competitors, surface technical constraints, and preserve sources for later decisions.
RESEARCH.md
RESEARCH.md - 02
Define the product
Turn evidence into users, scope, requirements, acceptance criteria, and a launch definition the build can follow.
PRD.md
PRD.md - 03
Design the system
Choose the stack, data model, APIs, security boundaries, deployment shape, and implementation constraints.
TECH_DESIGN.md
TECH_DESIGN.md - 04
Plan the build
Sequence the work file by file with phases, dependencies, verification commands, and completion criteria.
BUILD_PLAN.md
BUILD_PLAN.md - 05
Package the handoff
Generate agent instructions and export the complete context package for Cursor, Claude Code, Codex, or your team.
AGENTS.md + ZIP
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.
Claude Fable 5
Claude Opus 4.8
GPT-5.6 Sol
GPT-5.6 Terra
Gemini 3.1 Pro
Grok 4.5
DeepSeek V4 Pro
GLM 5.2
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.