Open-source runtime · MIT

One app, for people and for AI.

Your customers are starting to book, order and ask for help through ChatGPT, Claude and WhatsApp. With Copilotz you write your app once, and it works for them everywhere: the same actions, the same rules, one record of who did what for whom.

Get started View on GitHub

npm install copilotz@next

Building with an AI coding agent? Give it this first:

Read https://copilotz.com/AGENTS.md and follow it to build with Copilotz.

One app, 3 callers

Ana, on your web pageclicks “Book”
Your agent, working for Ana“book Wed 15:00”
An assistant, over MCPChatGPT, Claude

booking.makesame rules

One record

Who	For whom	Where
Ana	Ana	Web page
Your agent	Ana	WhatsApp
ChatGPT	Ana	MCP

Every write names who made it, for whom, and where.

Four ways in

Learn it, build with it, use it, or have it built.

Start wherever you are. Each one makes the next easier.

And what anyone builds can be shared on the Hub, so the next person starts further ahead.Browse the Hub

The runtime

Your app has new callers. They don't click.

People now ask an assistant to do things for them. Today that means a bot, an MCP server and a ChatGPT app beside your product, each with its own idea of who's asking and what they may touch. Copilotz makes them one app.

The same actions
A person, your own agent and someone else's call the same Action. You write it once.
The same rules
Access is checked at the data, on every read and write, whatever the door.
One record
Every write is an event that names who acted, for whom, and where.

Write the app once.

An app is a list of plugins built from 6 primitives. Your agents are one more client of it, not its center.

One View per medium: React on the web, text in a model's prompt, a widget inside ChatGPT.
Start in memory, keep it in SQLite, grow onto PostgreSQL by changing one line.
Runs on Node, Deno and Bun, built on web standards.
Test it all with a scripted model, no API calls.

Read the booking app, line by line

salon.ts

const salon = definePlugin({  name: "salon",  actions: {    "booking.make": defineAction({      input: booking,      // Asked of every caller, tools included.      allow: peopleOrContacts,      run: (ctx, input) =>        ctx.db.collection("booking").insertOne(input),    }),  },  // …the agent's tool calls the same Action  resources: {    agents: {      concierge: {        instructions: "Book slots.",        model: "openai/gpt-5.4-mini",        tools: ["book"],      },    },  },});

Where your app works

Web page
ChatGPT and Claude, over MCP
WhatsApp
Voice calls
REST API
Terminal
Your own agents

Compass

Your agents, with you and your team.

A workspace where agents work in your conversations and on your computer, remember what matters, and ask before running commands, changing files or publishing. It's built on Copilotz, so everything they do is on the record.

Start free See Compass

Free to start. Plans from $10 a month.

Launch plan

Launch team

You: Can we launch on Friday?
North
Asking East for a draft and South to check the numbers.
South
The pricing page says 12 a month; the plan says 10. Which is right?
North
Holding Friday until you confirm the price.
Waiting for you

Learn

Learn to build with agents by building a real one.

In 41 short lessons you build Compass yourself, from one agent and one message to a workspace for people and four agents, with every decision explained. The docs stay free; the course is the guided path.

The docs are free

The course

US$ 29one payment