Coddle Studio · uz · ru · en

Design your product with your AI agent

Spec, database schema, architecture and build plan — in one place. Your agent writes, Studio checks, connects and remembers

Works with Claude · Codex · any agent with MCP

Works with your stack and your agents

Claude Code
Codex
Cursor
MCP
PostgreSQL
Supabase
MySQL
Prisma
Drizzle

OpenAPI 3.1
Gherkin
Playwright
Docker Compose
Zod
TypeScript
Mermaid
DBML
AGENTS.md

From a conversation to a ready plan

Your agent works in Studio the way you do: through the spec, the schema and the architecture — and forgets nothing

01
Conversation
You describe the idea to your agent — it creates a project in Studio and writes the spec
I want online booking for a barbershop: a client picks a barber and a time, a barber sees their schedule
AgentCreated the «Barbershop» project and wrote the spec: roles, booking, reviews. Now the database schema. projects_create spec_write db_write
02
Spec
Roles, screens, data and acceptance criteria — Studio checks the spec against the schema
# Barbershop «Moustache»
Online booking with a barber and reviews after a visit
## Features
### Online booking
- A client sees free slots 14 days ahead
- A booking is confirmed in Telegram
- A barber sees their schedule
- Cancelling up to 2 hours before
Acceptance criteria4 / 4
03
Schema and data
Tables, a linter, migrations and real Postgres right in the browser
clients
iduuid
nametext
phonevarchar
masters
idint
nametext
appointments
idbigint
client_iduuid
master_idint
starts_attimestamptz
db_lint · 0
04
Architecture and build
Load simulation before launch, a build plan and AGENTS.md for the repository
client
load balancer
api ×3
postgres
p95
84 ms
Errors
0 %
Load
1 200 rps

From a conversation to a ready plan

Your agent works in Studio the way you do: through the spec, the schema and the architecture — and forgets nothing

coddle.uz/studio · Barbershop «Moustache»

I want online booking for a barbershop: a client picks a barber and a time, a barber sees their schedule

AgentCreated the «Barbershop» project and wrote the spec: roles, booking, reviews. Now the database schema. projects_create spec_write db_write

# Barbershop «Moustache»

Online booking with a barber and reviews after a visit

## Features

### Online booking

A client sees free slots 14 days ahead
A booking is confirmed in Telegram
A barber sees their schedule
Cancelling up to 2 hours before

Acceptance criteria4 / 4

clients

iduuid

nametext

phonevarchar

masters

idint

nametext

appointments

idbigint

client_iduuid

master_idint

starts_attimestamptz

db_lint · 0

client

load balancer

api ×3

postgres

p95

84 ms

Errors

0 %

Load

1 200 rps

MCP tools for your agent: 56
schema checks with one-click fixes: 14
export formats: from SQL and Prisma to OpenAPI and Docker Compose: 22
built-in plugins for the spec, the schema and the architecture: 12

Everything you need before the first line of code

Product spec

Roles, features, screens and acceptance criteria — a living document your agent reads and extends

Database schemas

DBML or the canvas, import from SQL and Prisma, a linter that fixes in one click

users

orders

Postgres sandbox

Test data, queries and migrations checked on real Postgres — nothing to install

select count(*) from orders;

→ 1 204 · 3 ms

Architecture and load

Build the system from blocks and see whether it survives the peak before launch

api ×3

cache

p95 84 ms

Versions and migrations

Every change is a version, a structural diff and a ready SQL migration

v3 → v4migration.sql

+ orders.status order_status

+ create index on orders(user_id)

− orders.state text

Teamwork

Invite links, shared access for your organization and public projects

AMDS

Plugins

Turn one on, and the document does more

Tests from the criteria, OpenAPI from the tables, a failure drill and Docker Compose from the architecture. A plugin adds a tab, checks, export formats and a tool for the agent.

Built-in only: no third-party code runs

Spec

Test cases

Tests from the acceptance criteria

Feature: Online booking
  Scenario: free slots
    Then 14 days ahead

Spec

Tasks for a tracker

Features and the plan as issues

gh issue create \
  --title 'Online booking'
  --label feature

Spec

Glossary

The words the spec relies on

## Glossary
- **Slot** — 30 minutes of a barber
- **Visit** — a booking that took place

Spec

Risks

What such products trip on

## Risks
- **Double booking** — a unique slot
- **Payment fails** — check again

Spec

Metrics

How to tell a feature works

## Metrics
- **Booking conversion** — above 30 %
- **Retention** — after 7 and 30 days

Spec

AGENTS.md to GitHub

A pull request with the project's context

branch  coddle/agents-md-20261014
commit  + AGENTS.md
PR      acme/shop#7

DB schema

REST API

OpenAPI 3.1 from the tables

GET    /clients
POST   /appointments
PATCH  /appointments/{id}

DB schema

Personal data

Where the schema keeps data about people

clients.phone     contact
clients.password  ⚠ hash
masters.passport  special

DB schema

Index advice

How the app will read the tables

+ users(email) [unique]
+ orders(user_id, created_at)
− orders(user_id)

DB schema

Naming rules

One convention for the whole schema

user       → users
firstName  → first_name
active     → is_active

Architecture

Failure drill

What if a block goes down

lb       outage    ✕
api ×3   ok        ✓
db       degraded  ~

Architecture

Docker Compose

The architecture on your laptop

services:
  db:  postgres:17-alpine
  api: { replicas: 3 }

Your agent already knows how to work with Studio

Studio is an MCP server: your agent reads and edits projects right from the chat, only within the permissions you grant

61 tools: the spec, the schema, migrations, simulation, plugins, discussions, the community and export
Sign-in through the browser and scoped permissions — revoke access at any time
AGENTS.md brings the whole project into the repository — no need to explain it again

coddle · MCP — Call log

projects_createbarbershop
spec_writespec
db_writeschema
db_lint5 → 0
db_migrationv1 → draft
arch_simulate×4 peak
spec_exportAGENTS.md

7 calls · the agent is workingno errors

See how it works

The agent writes the schema, the canvas draws it, Studio checks and fixes it

From the spec the agent designs the database and writes it straight into Studio — the canvas updates live.

Your agentcoddle · MCP

The agent is connected to Coddle over MCP and sees your projects.

Message your agent…

coddle.uz/studio/you/barbershop/schema

Studio/barbershop/SchemaDB schema

savedliveempty document

// the DBML will appear here

Lint · 0Versions · 0

The checks appear with the schema.

CanvasSandbox

The canvas is empty — the agent will draw the schema.

0:00 / 0:17

Also in Coddle

Coding practice for those who learn and those who teach

All about practice

TasksPython, JavaScript, C++ and Java, judged on hidden tests Battles1v1 duels in real time LeaderboardRating, streaks and achievements ClassesFor teachers: groups, lessons and homework

Start a project in a minute

Create a project and connect your agent — the first spec will be ready before your coffee cools

Design your product with your AI agent

From a conversation to a ready plan

Conversation

Spec

Schema and data

Architecture and build

From a conversation to a ready plan

Everything you need before the first line of code

Product spec

Database schemas

Postgres sandbox

Architecture and load

Versions and migrations

Teamwork

Turn one on, and the document does more

Test cases

Tasks for a tracker

Glossary

Risks

Metrics

AGENTS.md to GitHub

REST API

Personal data

Index advice

Naming rules

Failure drill

Docker Compose

Your agent already knows how to work with Studio

See how it works

Also in Coddle

Start a project in a minute