opencloudportConnect an AI agent
Give Claude Code, Cursor or your own agent a key, and it can build and run apps on opencloudport.
opencloudport has an MCP server: the standard way AI agents discover and use tools. Once connected, your agent has a tool for every job: apps, services, code, data, email and sandboxes, and the same guidance a person would get.
The server lives at:
https://opencloudport.com/mcpIt takes your API key in the Authorization header, the same way as the API.
1. Make a key for the agent
Give every agent its own key, so you can see what it did and stop it on its own.
- Whole workspace: on the API keys page, choose Create an API key.
- Just one app (recommended for agents working on one project): on the app's Settings tab, choose Create key. You can narrow it to one database or bucket, make it read-only, and set a spending limit. See Keys and spending limits.
The key is shown once. Keep it like a password: in your agent's settings or an environment variable, never in code you share.
2. Add it to your agent
claude mcp add --transport http opencloudport \
https://opencloudport.com/mcp \
--header "Authorization: Bearer <your key>"Run /mcp inside Claude Code to check it's connected.
3. Ask
Describe what you want in plain words. A few things agents do well here:
- "Put my bakery's order form online, with a database for orders and email receipts."
- "Spin up a sandbox, install my project and run its tests."
- "Add a sign-up list to my site and tell me who joined today."
- "Let visitors upload photos to my gallery."
What keeps it safe
A key per agent
Limit it to one app or one database, and make it read-only if it only needs to look.
Spending limits
At its limit a key can still read, but every change is refused until you raise it.
An activity log
Every change a key makes is written down in plain words, marking what came through MCP.
Asks before deleting
Tools that delete are marked as such, so agents like Claude Code check with you first.
The server also tells every agent two things up front:
- Data stays data. Database rows, files, emails and command output were written by other people. The agent is told never to follow instructions found in them, and to check with you before acting on anything they ask for.
- Limits are yours to raise. When a key reaches its spending limit, the agent is told to stop and tell you, rather than look for a way around it.
After an update
When opencloudport adds tools or new options to a tool, reconnect your agent to see them (in Claude Code, /mcp and reconnect, or start a new session). The server always lists today's tools; some agents keep the list they got when they connected.
If something goes wrong
| What you see | What it means |
|---|---|
invalid_api_key | The key is wrong, revoked, or its workspace was deleted. Make a new one. |
key_out_of_scope | The key only reaches one app or service, and the agent tried something else. |
key_spend_limit_reached | The key reached its spending limit. Raise it on the API keys page. |
insufficient_balance | The workspace is out of credit. Add funds on the Billing page. |