Archboard

npx add-mcp 'https://getarchboard.app/mcp'

Connect your AI agent to your diagrams through MCP

Archboard runs a remote MCP server. Add it to your AI app, sign in with Google or GitHub, and choose what the agent can use. Its edits appear live next to your team, its cursor shows who it works for, and you can undo your agent’s change in one step.

Watch an agent work next to your team

A scripted demo on the Archboard canvas, with sample names. The agent’s cursor shows who it works for, new shapes appear one by one, and a notice lets you undo your agent’s change in one step.

1. An AI agent working for Maya adds a database, its storage, and a cache next to the checkout API. Its cursor shows who it works for, and the new shapes appear one by one.
2. At the same time, Leo changes the checkout API from 3 to 5 replicas.
3. Maya clicks Undo in the notice: the agent’s change is taken back in one step, and Leo’s edit stays.

Connect in three steps

There’s no API key to paste: your AI app signs in to Archboard with OAuth, and you approve what it can use.

• MCP server URL: https://getarchboard.app/mcp

1. Add the server to your AI app: In your AI app’s connector or MCP settings, add Archboard as a remote MCP server with this URL. Some apps call it a custom connector.

2. Sign in and choose access: Your app opens Archboard. Sign in with Google or GitHub, pick the organizations and workspaces it can use, choose read, edit, or delete for each, and give it a name.

3. Ask for a diagram: Ask your agent to draw or update a diagram, then open it in Archboard to watch the changes arrive live.

   Try asking:

   • Draw our checkout service on Kubernetes with Postgres and Redis.
   • Add a cache next to checkout-api in my Payments diagram.
   • Undo your last change.

Your AI app needs to support remote MCP servers with OAuth sign-in.

What your agent does in Archboard

Edits land live

Your agent’s changes appear on the canvas as they happen, next to your team’s cursors. New shapes appear one by one as its cursor moves, and you can follow it around the canvas.

Labeled with who it works for

Its cursor shows the name you gave the connection and the person it works for, and the same name appears in change notices. Anyone watching through a public link sees the label too.

Undo your agent’s change in one step

Right after your agent changes a diagram, a notice lets you undo your agent’s change in one step. You can also ask the agent to revert one of its own changes within 7 days. A field someone changed afterward is kept.

The node packs you use

Agents add nodes from the same packs you use, so the result is a structured diagram you can keep editing.

Images, cleaned on the way in

With edit permission, your agent can add an image from a public https address, or a file from your computer in apps that can run commands. Archboard keeps only the processed image. Image metadata (location, camera, comments) is always removed.

You decide what it can reach

Choose what it can reach

When you approve an app, allow all your organizations and workspaces, including ones you join later, or only the ones you pick. Then choose read, edit, or delete for each. Delete only moves diagrams to the trash, where you can restore them.

Change or remove access anytime

Rename a connection, change what it can use, or remove it under Settings → Connected apps. Changes apply to its next call, with no reconnect.

Never more than you can do

An agent can’t do more than you can: your role and your plan limit it too. Apps that registered themselves are marked as unverified before you approve them.

A record you can check

Agent changes keep the app’s name in version history and in your organization’s audit log, which owners and admins can review on every plan.

On every plan, including Free

MCP works on every plan, including Free, with the same plan limits as the editor. Agent tool calls also have a per-minute rate limit. On Free you edit live with your AI agents; editing live with teammates needs Pro or Enterprise seats.

View pricing

Questions about AI agents

Which AI apps can connect?

AI apps that can add a remote MCP server with OAuth sign-in. Apps that require a client secret can’t connect. After you approve, Archboard sends you back to the app through an https address or a local (loopback) address on your computer. An app that returns through its own link scheme can finish connecting when Archboard allows that exact link; the developer guide lists them.

Do I need an API key?

No. Apps connect with OAuth 2.1 through Archboard’s own authorization server: you sign in with Google or GitHub and approve the app. There are no API keys to create, paste, or rotate.

Can an agent reach diagrams I didn’t allow?

No. It can use only the organizations and workspaces you allowed, and in each only what the permission there and your role allow. A call outside them is refused.

Can an agent delete my diagrams?

Only where you gave it Delete permission, and only by moving a diagram to the trash, where you can restore it. An agent can’t delete a diagram permanently.

How do I disconnect an agent?

Open Settings → Connected apps and remove its access. It is disconnected right away, and the changes it already made stay in your diagrams, with the app’s name in version history.

Built on open standards

A remote MCP server over Streamable HTTP, with OAuth 2.1 and PKCE from Archboard’s own authorization server, Client ID Metadata Documents or dynamic client registration, and protected resource metadata. Diagrams use a public, versioned JSON format with published JSON Schemas.

• Developer guide (English)
• Document format (English)

Machine-readable

• llms.txt
• llms-full.txt
• API catalog
• ARD catalog
• auth.md: OAuth 2.1 for AI agents (MCP); people sign in with Google or GitHub
• MCP server guide · Document format