Skip to content
Vibe(pre)Coding
enid
Sign inStart a project
Load your context before you build

Vibe (pre)Coding

Answer a few plain-language questions. Walk away with the AGENTS.md, PLAN.md and PRD.md your AI agent needs — before it writes a single line.

Start a projectSee how it works
Ready check4 / 4 equipped

  • AGENTS.md

    Rules your AI agent reads first

    Ready

  • PLAN.md

    The build order, step by step

    Ready

  • PRD.md

    What the product must do

    Ready

  • SECURITY.md

    What to check before launch

    Ready
Two ways to start

Skip this, and your AI agent is guessing.

Not ready

Vibe coding on a blank prompt

  • You re-explain the project in every new chat
  • The agent guesses your tech stack — sometimes wrong
  • Security is whatever the model remembers to add
  • One giant prompt trying to do everything at once
Ready

Vibe coding with your context loaded

  • AGENTS.md means the agent already knows the rules
  • Every decision tagged: your call, our recommendation, or unknown
  • SECURITY.md ships from the first prompt, not after a breach
  • PLAN.md breaks it into steps an agent can actually execute
See the difference

Same idea. Two very different starting points.

Just a prompt

> "build me a booking app for my small business"

  • · Framework — guessed
  • · Database — guessed
  • · Security — whatever it remembers

The agent starts writing code immediately, and hopes it guessed right.

Vibe(pre)Coding

> "build me a booking app for my small business"

  • Understood: booking app, small business
  • 3 quick questions answered
  • AGENTS.md · PLAN.md · PRD.md · SECURITY.md ready

Now the agent has everything it needs — before it writes a line.

How it works

You bring the idea. We handle the engineering paperwork.

AI coding agents work far better when they have real project context. Writing that context normally means knowing what a PRD, an architecture doc and an AGENTS.md are. Here, you don't.

Phase 01

Describe it in your own words

One box, no jargon. Say what you want to build the way you'd explain it to a friend.

Phase 02

We read it back to you

You confirm what we understood, or correct it in plain language. Nothing gets built on a misunderstanding.

Phase 03

A few simple questions

Only the ones that matter. "I'm not sure" is always an option, and we'll recommend something.

Phase 04

Files your coding agent can use

Plans, rules and context, written to be acted on. Download them as a folder and start building.

Nothing invented

We keep what you said apart from what we recommended and what we assumed. If something is unknown, it says so.

Change it by talking

"Add team collaboration." We work out which files that touches, show you, and update them.

Security isn't an afterthought

This isn't just about the project you asked for.

Most people don't know to ask for security — by the time they find out, a key is already in the git history. So SECURITY.md ships with every project, however small, whether you asked for it or not.

SECURITY.md

In every project

  • Secrets & environment variables
  • Input validation
  • Permission checks
  • Dependency review
  • Errors & logs

Added automatically when your project needs it

  • File uploads
  • Payments
  • Multi-tenancy
  • Prompt injection

A pre-launch checklist

A rotate-first plan for a leaked key

Audit questions you can paste straight to your coding agent

“Show me every query that looks up a record by id without also filtering by owner.”
Your loadout

What you walk away with

A folder of files, each one explained in plain English. Only the files your project actually needs.

  • AGENTS.mdInstructions for AI coding agents working on your project.Ready
  • DESIGN.mdA real visual direction, so the result doesn't look like a generic template.Ready
  • PROJECT.mdAn overview of what you're building.Ready
  • PRD.mdA detailed description of what the product should do.Ready
  • PLAN.mdA step-by-step implementation plan.Ready
  • ARCHITECTURE.mdHow the different parts fit together.Ready
  • SECURITY.mdThe security rules for your project, and what to check before launch.Ready
  • TASKS.mdThe work that needs to be completed.Ready
Project dashboard

Every project gets its own command deck

Once the files are generated, the project page stops being a wizard and becomes mission control. Five views in one sidebar — the whole build lives here.

Command deck5 views

  • Home

    The build at a glance — progress since the first prompt

  • Steps

    The checklist that takes you from idea to generated files

  • Task list

    The board — every task, from backlog to done

  • Connect

    The link your coding agent plugs into

  • How to use

    Four steps to a fully automated build

The task list

Your plan becomes a board the agent can work

Every step in PLAN.md and TASKS.md turns into a card across six columns. Drag it, nudge it with arrows — the board holds the rules, so nothing skips straight to Done.

Task list
6 columns

Backlog

Design tokens

P2

To do

Set up auth

P1

In progress

Booking form

P0

Testing

Booking flow e2e

P1

Done

Project scaffold

P2

Cancel

Dealt straight from your plan — no manual entry.

Forward one column at a time, back any distance. The board enforces it.

Every card: urgency P0–P5, type, labels, start and due dates.

MCP

Connect your agent. Watch it build.

In the Connect tab, create a link and paste it into opencode, Claude Code, or Cursor. Then paste the kickoff prompt — your agent reads the files, works the plan, and keeps the board in sync while it builds.

Connect — one time

$ claude mcp add --transport http vibe-planner <url>

  • Works with opencode, Claude Code and Cursor
  • The URL is shown once, then never again
  • Revoke it and the agent is locked out instantly

One URL per project. No keys to manage, nothing to configure.

Kickoff — paste and go

> paste the kickoff prompt

  • Reads AGENTS.md → PROJECT.md → PLAN.md, in order
  • Works the plan top-to-bottom — no skipping steps
  • Moves the cards on your board as it builds

The prompt is written for you, per project — with the agent instructions baked in.

8 tools your agent gets over MCP

  • vibe_get_project
  • vibe_list_files
  • vibe_get_file
  • vibe_get_context
  • vibe_list_tasks
  • vibe_create_task
  • vibe_update_task
  • vibe_delete_task

Tokens are shown once at creation — copy them then.

Stored as SHA-256 hashes; even we can't read them back.

Revoking works instantly, per project.

“You paste one prompt. The agent reads the brief, builds the plan, and the cards make their own way to Done.”

You have an idea. That's the hard part.

Everything after it is paperwork. Let's get it done in a few minutes.

Start a project