Your applications in your AI assistant
Connect Claude, ChatGPT, Claude Code, Cursor, VS Code or Codex to your Square Cloud account. Ask for status, logs and metrics, restart, edit files and create resources without leaving the conversation. Before deleting or restoring, Square Cloud takes a safety snapshot.
https://mcp.squarecloud.app/mcp
In Settings, Connectors, add a custom connector with this address. The first time, Claude opens the Square Cloud authorization.
No key to copy: you choose the permissions on the authorization screen.
Available on every plan, from Hobby up. See plans
See MCP in action
Three example conversations with an assistant connected to your account. Pick one and follow each tool it uses.
The Discord bot went down again. Can you see what happened?
> node index.jsLogged in as SquareBot#4821TypeError: Cannot read properties of null (reading 'send') at Object.execute (commands/ping.js:12:27)
The /ping command tries to send to the message's channel, which comes empty in direct messages, and the bot stops with that error. I'll reply through the interaction itself and restart.
discord-botRunningDone. The bot is back and /ping now answers in direct messages too.
Reads the logs, finds the failing line, fixes the file and restarts.
What your assistant can do
It works with the same applications, databases and files you see in the dashboard, within the permissions you granted.
Monitor
Status, logs, metrics and network of each application, so you can understand an error without leaving the conversation.
Operate
Start, stop and restart, read and edit files, send a new version of the code, and create, change or remove environment variables.
Create
New applications (including from a project on your computer), databases and snapshots, plus the custom domain of each application.
You stay in control
Critical actions ask for confirmation in your AI client, and deleting or restoring takes a safety snapshot first.
- The client asks for permissionEach tool tells your AI client whether it only reads or can destroy something. Before an action like that, it asks for permission in the conversation itself.
- Snapshot before actingDeleting an application or database and restoring a snapshot take a safety snapshot first, or reuse one made in the last few minutes.
- 3The snapshot stays in the dashboardAfter a delete, the snapshot stays in the dashboard: download it and upload it as a new application if you need to.
- If the snapshot is still processing, nothing runs and the client tries again in about 2 minutes.
- If your plan's daily snapshot limit is already used up, or if the snapshot fails, the action runs anyway, without a snapshot.
Secrets stay out of the model
Variable values, .env files and keys never reach the model. When reading, it only sees variable names.
Permissions you choose
On the authorization screen, untick what you don't want to grant. In code tools, the key can be read only.
Everything in account activity
What the assistant does shows up in account activity with the MCP badge. To disconnect, revoke the key in account security.
Limits per account
Limits stop an assistant stuck in a loop. Each call also counts toward the account's API limit.
Tool calls per minute
- Hobby 1 GB
- 15
- Hobby 2 GB
- 30
- Standard and up
- 60
File operations per minute, on any plan
- Reads
- 20
- Writes
- 10
- Deletions
- 5
- Searches
- 3
On every plan, tools use up to half of the API limit per minute, at most 60.
A plan for every stage of your project
Start small and scale when you need to, with no loyalty contracts.
Hobby
To validate ideas, bots and personal projects
From R$12.49/mo ≈ $2.25
- 1 GB RAM
- 1 vCPU
- Up to 4 projects
- Scales up to 2 GB on the same plan
Standard
RecommendedFor production, with databases and custom domains
From R$44.99/mo ≈ $8.10
- 4 GB RAM
- 4 vCPU
- Up to 16 projects
- Scales up to 8 GB on the same plan
Pro
For high scale, with daily application snapshots and advanced analytics
From R$124.99/mo ≈ $22.50
- 12 GB RAM
- 6 vCPU
- Up to 48 projects
- Scales up to 24 GB on the same plan
4.9 out of 5 · 402 reviews on Google + TrustpilotTrustpilot
MCP frequently asked questions
What is MCP?
The Model Context Protocol is an open standard for AI assistants to use external tools. The Square Cloud MCP server gives your assistant tools to view and operate the applications and databases in your account.
Which plan do I need?
Any plan, from Hobby up. Without an active plan, the server rejects calls. The per-minute call limit follows the plan.
Which clients work?
Claude (claude.ai and apps) and ChatGPT connect with the server address. Claude Code, Cursor, VS Code and Codex connect with an API key, with the command ready on the dashboard's tools page.
Does the model see the values of my environment variables?
No. Variable values, .env files and keys never reach the model. When reading, it only sees the names.
What can't be done through MCP?
Downloading snapshots and certificates and viewing the account activity history. That stays in the dashboard only. Projects larger than 1 MB zipped go through the CLI or the dashboard.
How do I disconnect an assistant?
In account security, revoke its key. Claude and ChatGPT appear there as connector keys, and revocation takes effect immediately.