Apps and deploys
Create, publish, restart, and inspect apps, logs, metrics, runtimes, domains, and deployment history.
Model Context Protocol
Your infrastructure speaks to your assistant.
Vertra's official MCP server gives AI agents real tools to deploy apps, read logs, manage databases, and restore snapshots — within the scopes you grant.
Your assistant
MCP
Vertra Cloud
apps:readdatabases:writesnapshots:read
Open sourcepublic, auditable implementation
MITa simple license to use and contribute
Local + webstdio through npx or remote connector
Scopedyou decide what each agent can access
How it works
Your intent becomes an authorized Vertra call.
MCP translates natural-language requests into tools with structured inputs and responses. Access rules remain the same as the API.
Choose access
Create a key with only the scopes you need or authorize the web connector through Vertra. Read and write stay separate.
The client finds the tools
Claude, Cursor, VS Code, and other compatible clients receive Vertra's catalog and know which operations they can call.
The assistant acts and explains
Each tool calls the public API, returns a structured response, and lets the agent continue with the real result.
Tools
From first deploy to daily operations.
One connector covers the resources you already manage in Vertra.
Databases
Create and operate PostgreSQL, MySQL, MongoDB, and Redis with status, metrics, and credentials when the scope allows it.
Files and environment
Read and change files, upload a local folder, and manage variables without exposing secret values in listings.
Workspaces and account
Inspect usage, members, roles, activities, notifications, and shared team resources.
Operations
Follow deployments, logs, and resource status without leaving your workflow.
Snapshots and operations
Create, restore, and download snapshots, purge cache, and monitor resource health without leaving your agent.
Two ways to connect
In the browser or on your computer.
Use remote when you want to sign in with your account. Use the local package when the agent also needs to work with your project files.
Remote connector
Add the URL to your client, sign in to Vertra, and confirm the scopes. No package install or API key copied into the browser.
https://mcp.vertracloud.app/mcpLocal package
Run through stdio with Node 18 or newer and provide VERTRA_API_KEY to the process. This mode enables local-file operations.
Access control
The agent gets only what you grant.
MCP does not bypass platform protections. It uses the public API and respects the key, scopes, and account that authorized the connection.
Least privilege
Start with read access and add write only where the workflow truly needs it.
Protected secrets
Listing variables shows names, never values. Sensitive operations require dedicated scopes.
Revocable access
Revoke the key or Vertra authorization to cut access without reconfiguring your infrastructure.
Open source
Thin by design. Public by trust.
The @vertracloud/mcp project is a small layer over Vertra's public API: tools, stdio transport, and a reusable server. No database, cache, or credential of its own. Read the code, open an issue, or contribute.
Open repositoryFrequently asked questions
Before you connect.
The essentials about compatibility, billing, and access.
Does MCP cost anything beyond my plan?
Using the MCP server does not create a separate subscription. Actions follow the same rules and limits of the Vertra feature you already use.
Which AI client can I use?
Any client compatible with Model Context Protocol. The local package works over stdio; the remote connector uses Streamable HTTP and authorization through your Vertra account.
Can the assistant see my environment variables?
The listing tool returns names only. Setting or removing a variable requires its own scope, and a submitted value is never echoed in the response.
Why are some tools local only?
Creating an app from a folder, uploading files, and downloading snapshots need to read or write on your computer. A remote browser server does not have that access — and should not.
Can I use it with a workspace?
Yes. Workspace tools can inspect resources, members, roles and usage according to your account permissions and granted scopes.
Start with read access
Give your assistant real context.
Create a key with minimal scopes, connect your client, and ask it to list your apps. Then expand access only where it makes sense.
Create API key