Skip to main content
Gcore MCP Server provides approximately 400 tools for managing Cloud infrastructure—virtual machines, storage, networking, and security—through natural language. Configuration and client setup are covered in the Gcore MCP Server overview.

Configure for Cloud

The GCORE_TOOLS environment variable controls which tools are exposed. Setting GCORE_TOOLS="cloud.*" loads all Cloud tools, while predefined toolsets provide focused subsets for specific workflows.
GCORE_TOOLS="cloud.*"
Predefined toolsets group related operations: instances for virtual machine operations, storage for volumes and file shares, networking for networks, floating IPs, and load balancers, and security for security groups and SSH keys. Combining toolsets with GCORE_TOOLS="management,instances,storage,networking" loads tools from multiple categories. Cloud tools follow the cloud.<resource>.<operation> naming pattern. Compute tools include instances for virtual machines with list, create, get, delete, start, stop, and resize operations, baremetal for bare metal servers with power management, and gpu_baremetal and gpu_virtual for GPU cluster management. Storage tools include volumes for block storage with attach and detach operations, volume_snapshots for point-in-time copies, and file_shares for shared file systems. Networking tools include networks for virtual networks, floating_ips for public IP addresses with assign and unassign operations, load_balancers for traffic distribution, and security_groups for firewall rules. Management tools include projects, regions, quotas, and ssh_keys.

Cloud operations

Natural language requests trigger the corresponding API operations. Listing projects requires no parameters—a request like “List all my cloud projects” returns project names and IDs. Instance creation requires project ID, region, flavor specifications, and image selection. Requesting “Create a virtual machine in project 780356 with 4 vCPUs, 8 GB RAM, Ubuntu 22.04, and 80 GB SSD in Luxembourg region” provisions an instance with the specified configuration. The instance typically becomes available within two minutes. Instance management includes lifecycle operations and resource changes. A request to “Stop instance 12345” powers down the virtual machine while preserving its configuration. Resizing with “Resize instance 12345 to 8 vCPUs and 16 GB RAM” changes the instance flavor—this operation requires a stopped instance. Storage operations span volume lifecycle and attachment. Creating a volume with “Create a 100 GB SSD volume named data-volume in project 780356, Luxembourg region” provisions block storage. Attaching with “Attach volume data-volume to instance 12345” makes the volume available to the instance. Snapshots capture volume state—“Create a snapshot of volume data-volume” creates a point-in-time copy. Networking operations configure connectivity and access. Listing floating IPs shows available public addresses. Creating and assigning with “Create a floating IP and assign it to instance 12345” provisions a public IP and associates it with the instance. Load balancer creation requires backend instances and port configuration—“Create a load balancer for instances 12345 and 12346 on port 80” distributes traffic across the specified instances. Security group operations define firewall rules. Creating a group with “Create a security group that allows SSH (port 22) and HTTP (port 80) from anywhere” provisions a rule set with the specified ingress permissions. Adding rules with “Add a rule to security group 5678 allowing HTTPS (port 443)” extends an existing group.

Multi-step workflows

Complex deployments involve multiple tools executed in sequence. Deploying a web server requires creating a VM, provisioning a floating IP, configuring security rules, and applying the security group. A combined request like “In project 780356, Luxembourg region: create a VM with 2 vCPUs, 4 GB RAM, Ubuntu 22.04; create a floating IP and assign it; create a security group allowing SSH, HTTP, and HTTPS; apply the security group to the VM” triggers the appropriate tool sequence. Database server setup combines compute, storage, and security configuration. Requesting “Create a database server setup with 4 vCPUs, 16 GB RAM, a 500 GB SSD data volume, private network without floating IP, and a security group allowing PostgreSQL port 5432 from the internal network” provisions the complete stack. Resource cleanup queries identify unused assets. “List all unattached volumes in project 780356” shows volumes not connected to instances. “Find floating IPs that are not assigned to any instance” identifies billable IPs without associated workloads.

Default project and region

Environment variables can set default context for Cloud operations, eliminating the need to specify project and region in every request.
GCORE_CLOUD_PROJECT_ID="780356"
GCORE_CLOUD_REGION_ID="76"
With these defaults, requests like “Create a VM with 2 vCPUs and 4 GB RAM” use the specified project and region automatically.
InfoTool parameters and behavior reflect the Gcore Cloud API. Volume management is documented in the block storage guide. Load balancer configuration is covered in the load balancer guide. Tool filtering options are available in the GCORE_TOOLS reference.