glesys-mcp

MCP server for the Glesys API. Manage your Glesys infrastructure from Claude Code, Claude Desktop, Cursor, VS Code or any other MCP client.

Examples

Once connected, you talk to your client in plain language and it picks the right tools:

• "We just hit the front page of Hacker News. Double web01's CPU and memory, and tell me what the panic costs per month."
• "I'm demoing in an hour. Postgres with an app user, a Debian server in Falkenberg allowlisted to reach it, and the connection string in my hand."
• "New project, new domain. Register example-app.se if it's available, create a small Ubuntu server in Falkenberg, and point the domain at the server's IP."
• "web01 is on fire. Pull it out of the load balancer, hard reset it, and don't put it back until it responds."
• "Someone deleted DNS records for the docs site. Check the zone, tell me what's missing, and put it back."
• "Set up the whole stack for the new app: two web servers in Stockholm, a file storage volume both can mount, a MySQL they're allowed to reach, and a load balancer in front with both servers as targets."
• "Every IP we own should have a PTR record matching its server's hostname. Find the ones that don't and fix them."
• "The new developer starts Monday. VPN user, SSH key on the staging servers."
• "Give me a server that's ready when it boots: Debian, a deploy user with my SSH key, and Docker installed via cloud-init. Show me the cloud-config before creating anything."
• "Is it us or is it Glesys? Check the status page, and if Glesys looks fine, pull the load balancer error log."

Every tool is annotated as a read or a write, so your client can decide what to confirm. Claude Code runs reads directly and asks before writes; other clients, such as Claude Desktop, may ask before every call until you allow the tool.

Setup

Requires Node.js 22 or later. Your client starts the server with npx, which downloads @glesys/mcp from npm on first use, so there is nothing to install.

1. Create an API key

Create an API key for your project in the Glesys Control Panel. New keys start fully restricted:

• Allowed hosts: add the public IP of the machine running the MCP server.
• Allowed functions: allow the modules you want to use (server, serverdisk, sshkey, ip, networkadapter, network, privatenetwork, loadbalancer, domain, email, objectstorage, filestorage, blockstorage, archive, database, vpn, colocationrack, remotehands, shipment).

If every tool returns a 403 or "not allowed" error, it is one of these two.

2. Store the key in your shell

Skip this step if you only use Claude Desktop.

Keep the key in your shell profile rather than in your client's config file. Config files are easy to share or commit by accident, and commands like claude mcp get print them in full. Depending on your system:

• zsh (the macOS default): add export GLESYS_API_KEY='your-api-key' to ~/.zshrc.
• bash: add the same line to ~/.bashrc on Linux or ~/.bash_profile on macOS.
• Windows: run setx GLESYS_API_KEY "your-api-key" in a terminal. In WSL, use the zsh or bash line instead.

Then quit your terminal and your client completely (not just the window) and reopen them, since they read the variable only at startup. Claude Desktop is the exception: it only passes the server the variables in its config file, so there the key goes in the file as is.

3. Configure your client

The server reads two environment variables, plus an optional third:

• GLESYS_PROJECT_KEY: your project key, e.g. cl12345
• GLESYS_API_KEY: the API key
• GLESYS_ENABLE_DESTRUCTIVE: set to true to register the destructive tools (delete/destroy/remove/release, plus dns_generate_auth_code, since a leaked code lets anyone transfer the domain). Without it they do not exist as far as the client is concerned, which protects you even if your client skips confirmations. To enable them later, add the variable to the env block in your client config and restart the client (for Claude Code, run claude mcp remove glesys, then claude mcp add again with the extra --env).

Claude Code

claude mcp add glesys --scope user --env GLESYS_PROJECT_KEY=cl12345 --env GLESYS_API_KEY='${GLESYS_API_KEY}' -- npx -y @glesys/mcp

The single quotes matter: without them your shell expands the variable and the key is written to ~/.claude.json in plain text. In Windows CMD, leave out the single quotes; keep them in PowerShell.

--scope user makes the server available in all your projects. Without it, the server is only registered in the current directory. There is also --scope project, which shares the server with everyone in a repository via a committed .mcp.json. Keep the credentials as environment variable references there.

Claude Desktop (claude_desktop_config.json)

{
  "mcpServers": {
    "glesys": {
      "command": "npx",
      "args": ["-y", "@glesys/mcp"],
      "env": {
        "GLESYS_PROJECT_KEY": "cl12345",
        "GLESYS_API_KEY": "your-api-key"
      }
    }
  }
}

Claude Desktop has no documented way to reference environment variables, so the key goes in the file as is.

Cursor (~/.cursor/mcp.json)

{
  "mcpServers": {
    "glesys": {
      "command": "npx",
      "args": ["-y", "@glesys/mcp"],
      "env": {
        "GLESYS_PROJECT_KEY": "cl12345",
        "GLESYS_API_KEY": "${env:GLESYS_API_KEY}"
      }
    }
  }
}

VS Code (.vscode/mcp.json)

{
  "servers": {
    "glesys": {
      "command": "npx",
      "args": ["-y", "@glesys/mcp"],
      "env": {
        "GLESYS_PROJECT_KEY": "cl12345",
        "GLESYS_API_KEY": "${env:GLESYS_API_KEY}"
      }
    }
  }
}

Multiple projects

Each server handles one project. For more, add one entry per project, each with its own name and API key:

claude mcp add glesys-prod --scope user --env GLESYS_PROJECT_KEY=cl12345 --env GLESYS_API_KEY='${GLESYS_API_KEY_PROD}' -- npx -y @glesys/mcp
claude mcp add glesys-dev --scope user --env GLESYS_PROJECT_KEY=cl00000 --env GLESYS_API_KEY='${GLESYS_API_KEY_DEV}' --env GLESYS_ENABLE_DESTRUCTIVE=true -- npx -y @glesys/mcp

Name the project when you ask ("list the servers in dev"), and the client picks that server's tools. Since each entry has its own environment, you can enable destructive tools for one project and leave them off for another.

Tools

Each tool maps 1:1 to a Glesys API function (server_create calls POST /server/create). See the API documentation for details.

Every tool has MCP annotations (readOnlyHint/destructiveHint) so your client can ask before writes. Destructive tools are off by default; GLESYS_ENABLE_DESTRUCTIVE in the setup section enables them.

| Group | Tools | |---|---| | server_* | list, details, status, resourceusage, costs, create, edit, destroy, start, stop (soft/hard/reboot), reset, templates, listplans, allowedarguments, estimatedcost, previewcloudconfig, networkadapters, backup, estimated_backup_cost, list_backups, create/delete_manual_backup, create_from_backup | | serverdisk_* | limits, estimatedcost, create, reconfigure, updatename, delete | | sshkey_* | list, add, remove | | ip_* | list_own, list_free, details, take, release, add_to_server, remove_from_server, set_ptr, reset_ptr, estimated_cost | | networkadapter_* | create, details, edit, delete | | network_* | list, details, create, edit, delete | | privatenetwork_* | list, details, create, edit, delete, list_segments, create_segment, edit_segment, delete_segment, estimated_cost | | loadbalancer_* | list, details, create, edit, destroy, add/edit/remove_backend, add/edit/enable/disable/remove_target, add/edit/remove_frontend, list/add/remove_certificate(s), errors, add_to/remove_from_blocklist | | dns_* | list_domains, domain_details, add_domain, edit_domain, delete_domain, export_zone, list_records, add_record, update_record, delete_record, check_availability, pricelist, allowedarguments, register_domain, transfer_domain, renew_domain, set_autorenew, change_nameservers, generate_auth_code | | email_* | overview, list, quota, costs, createaccount, editaccount, resetpassword, createalias, editalias, delete | | objectstorage_* | list_instances, instance_details, create_instance, edit_instance, delete_instance, create_credential, delete_credential, estimated_cost, costs | | filestorage_* | list_volumes, volume_details, list_plans, create_volume, edit_volume, resource_usage, delete_volume | | blockstorage_* | list_volumes, volume_details, list_tiers, estimated_cost, create_volume, edit_volume, resize_volume, attach_volume, detach_volume, delete_volume | | archive_* | list, details, allowedarguments, create, resize, changepassword, changedescription, delete | | database_* | list, details, listplans, create, delete, connection_details, update_allowlist, estimated_cost, create/delete_database, create/delete_user, reset_user_password, set_user_permissions | | vpn_* | list_users, create_user, edit_user, delete_user | | colocationrack_* | list | | remotehands_* | list, create, cancel | | shipment_* | list, create, edit, cancel | | glesys_* | status |