Skip to content

AI Assistant Connector (MCP)

Mengi Cloud has a hosted MCP server, so you can connect your account to Claude and other AI assistants that support the Model Context Protocol. Once connected, you can ask the assistant to inspect and manage your infrastructure in plain language — list clusters, check why a deployment is unhealthy, query metrics and logs, create a database, rotate a secret.

The connector URL is:

https://mcp.mengi.cloud/mcp

You sign in with your normal Mengi Cloud account and approve access in the browser. There is no API key to create, copy or paste.

Add the connector

Claude (web and desktop)

  1. Open Settings → Connectors and choose Add custom connector.
  2. Enter the URL https://mcp.mengi.cloud/mcp.
  3. Click Connect. You will be sent to Mengi Cloud to sign in (if you are not already) and asked to approve access.
  4. Approve, and you are returned to Claude with the connector active.

Claude Code

Terminal window
claude mcp add --transport http mengi https://mcp.mengi.cloud/mcp

The first time you use it, Claude Code opens your browser to sign in and approve access.

Other MCP clients

Any client that supports remote MCP servers over HTTP with OAuth can connect. Point it at https://mcp.mengi.cloud/mcp; the client discovers the sign-in flow automatically. Clients that only support a static token can instead send a Mengi Cloud API key as a bearer token — see Using an API key instead.

What the assistant can do

The connector exposes your account’s functionality as tools, grouped roughly the way the dashboard is:

Area Examples
Clusters List and inspect clusters, create and delete them, check resource metrics, discover workloads, review and approve Kubernetes upgrades, import an existing cluster
Deployments Deploy a container image or Helm chart, update image or scale replicas, inspect live status when something is unhealthy, promote between clusters, back up and restore
Databases Create managed databases, resize storage, read connection credentials, take and restore backups
Secrets List, create, update and delete secrets, scoped to a deployment or organization
Observability Query metrics with PromQL and logs with LogQL, list firing alerts, create alert rules, silence an alert
Registry List images and tags in your private registry, manage registry credentials
Domains Manage DNS credentials, list zones and records for custom domains
Organizations List organizations and members, invite and remove members, transfer resources
Billing Review your subscription, transactions and current usage

A good way to start is simply asking “what have I got running on Mengi Cloud?” — the assistant will list your clusters and deployments and go from there.

What approving access grants

Approving the connector gives the assistant the same access your own account has, including reading secrets and database credentials, creating and deleting infrastructure, and viewing billing information. It does not grant platform-administration access, even if your account is a platform admin.

Two things are worth knowing before you connect:

  • Actions cost money. Creating clusters and databases provisions real cloud infrastructure and is billed like any other resource you create.
  • Some actions are irreversible. Deleting a cluster removes everything running on it; deleting a database removes its data.

Destructive tools are marked as such, so a well-behaved client asks you to confirm before running them. Treat those confirmations as real — read what is about to be deleted before approving.

Reviewing and revoking access

Access is granted per client application and lasts until it is revoked.

Go to Settings → Connected applications in the dashboard to see every application you have authorized, when you connected it and when it was last used. Click Revoke access to cut one off.

Revocation takes effect immediately — the application’s next request fails, rather than working until its credential expires. Use it without hesitation if a device is lost or you no longer trust a client.

You can also disconnect from the assistant’s own settings, which is the tidier option when you simply no longer want the connector. Revoking from the Mengi Cloud side is the one that guarantees access stops.

Revoking does not undo anything the assistant already did; it only prevents further access. Reconnecting later is just a matter of approving again.

Using an API key instead

For clients that do not support OAuth — or for scripted and headless use — you can authenticate with a Mengi Cloud API key by sending it as a bearer token:

Authorization: Bearer mc_your_api_key_here

See API — Getting Started for how to create a key. The same warnings apply: a key carries your account’s full permissions, so keep it out of version control and rotate it periodically.

Troubleshooting

The assistant says it cannot reach Mengi Cloud. Check that the URL is exactly https://mcp.mengi.cloud/mcp, including the trailing /mcp.

You are asked to sign in every time. Your client is not storing the returned credentials. Check that it supports remote MCP servers with OAuth rather than only local servers.

A tool reports a permission error. The connector can only do what your account can. If the resource belongs to an organization, check your role in Settings → Organizations — some actions require the admin or owner role.

A tool reports something is still provisioning. Creating clusters, databases and deployments is asynchronous. Ask the assistant to check the status again in a few minutes rather than retrying the creation, which would create a second resource.