GCP Compute Operations
MCP Servers
- Compute Engine:
https://compute.googleapis.com/mcp(Streamable HTTP) - Resource Manager:
https://cloudresourcemanager.googleapis.com/mcp(Streamable HTTP) - Auth: OAuth 2.0 via Google IAM — service account key (
GOOGLE_APPLICATION_CREDENTIALS) orgcloud auth application-default login - Requires:
GCP_PROJECT_IDenvironment variable
Key Difference from On-Prem MCP Servers
Google Cloud MCP servers are remote HTTP endpoints hosted by Google — nothing to install locally. Authentication uses OAuth 2.0 / Google IAM, not API keys. The transport is Streamable HTTP, not stdio.
Available Tools (29)
Resource Manager (1 tool)
| Tool | What It Does |
|---|---|
search_projects | Search for GCP projects by parent, ID, or filter. Use this first to discover available projects. |
Instance Management (8 tools)
| Tool | What It Does |
|---|---|
list_instances | List VM instances in a project (filter by zone) |
get_instance_basic_info | Get core instance details — name, status, machine type, network interfaces, IPs |
create_instance | Provision a new VM instance |
delete_instance | Remove a VM instance |
start_instance | Boot a stopped instance |
stop_instance | Halt a running instance |
reset_instance | Restart an instance (hard reset) |
set_instance_machine_type | Resize an instance (change machine type) |
Instance Groups (3 tools)
| Tool | What It Does |
|---|---|
list_instance_group_managers | List managed instance groups (MIGs) |
get_instance_group_manager_basic_info | Get MIG details — template, target size, autoscaling |
list_managed_instances | List instances within a MIG |
Instance Templates (3 tools)
| Tool | What It Does |
|---|---|
list_instance_templates | List available instance templates |
get_instance_template_basic_info | Get template basics — name, description |
get_instance_template_properties | Get template specs — machine type, disks, network interfaces |
Disks & Storage (5 tools)
| Tool | What It Does |
|---|---|
list_disks | List persistent disks in a project |
get_disk_basic_info | Get disk metadata — size, type, attached instances |
get_disk_performance_config | Get disk performance settings — IOPS, throughput |
list_snapshots | List disk snapshots |
list_instance_attached_disks | List disks attached to a specific instance |
Resource Discovery (3 tools)
| Tool | What It Does |
|---|---|
list_machine_types | List available machine types in a zone (e2, n2, c3, etc.) |
list_accelerator_types | List available GPU/TPU types in a zone |
list_images | List available boot images (Debian, Ubuntu, CentOS, etc.) |
Reservations & Commitments (6 tools)
| Tool | What It Does |
|---|---|
list_reservations | List capacity reservations |
get_reservation_basic_info | Get reservation details |
get_reservation_details | Get comprehensive reservation info |
list_commitments | List committed use discounts (CUDs) |
get_commitment_basic_info | Get commitment details — term, status |
list_commitment_reservations | List reservations tied to a commitment |
Operations (1 tool)
| Tool | What It Does |
|---|---|
get_zone_operation | Check status of an async operation (create, delete, resize) |
Workflow: GCP Infrastructure Audit
When a user asks "show me our GCP compute resources":
- Discover projects:
search_projectsto list all accessible GCP projects - List instances:
list_instancesper project — name, status, zone, machine type - Network interfaces:
get_instance_basic_infoto see IPs, VPC, subnet per VM - Instance groups:
list_instance_group_managersfor autoscaled workloads - Disk inventory:
list_disks— size, type, attached instances - Report: GCP compute inventory with status and network details
Workflow: VM Troubleshooting
When investigating a GCP VM issue:
- Find the VM:
list_instancesfiltered by name or zone - Get details:
get_instance_basic_info— status, machine type, network interfaces, external IP - Check disks:
list_instance_attached_disks— any missing or detached? - Check operation:
get_zone_operation— any pending/failed operations? - Cross-reference: Use
gcp-cloud-loggingfor instance logs,gcp-cloud-monitoringfor metrics - Report: VM health assessment with findings
Workflow: Capacity Planning
When planning GCP compute capacity:
- Current usage:
list_instances— count by machine type and zone - Reservations:
list_reservations— reserved capacity vs actual usage - Commitments:
list_commitments— CUD utilization and expiration - Available types:
list_machine_types— what's available in target zones - GPU availability:
list_accelerator_types— GPU/TPU options per zone - Report: Capacity utilization with recommendations
Important Rules
- Google Cloud MCP servers are remote — no local install, runs on Google's infrastructure
- OAuth 2.0 authentication — uses service account or user credentials via IAM
- Project-scoped — all operations are scoped to the configured GCP project
- Write operations available — unlike AWS Network MCP, Compute Engine MCP can create/modify/delete VMs. Use ServiceNow CR gating for changes.
- Record in GAIT — log all GCP operations for audit trail
Environment Variables
GCP_PROJECT_ID— Google Cloud project IDGOOGLE_APPLICATION_CREDENTIALS— Path to service account key JSON file (or usegcloud auth application-default login)
Failure Behavior
- On a tool error (timeout, unreachable host, malformed response), report the failure and its error message directly to the user rather than fabricating or guessing at results.
- Do not automatically retry a write/mutating operation after a failure — surface the error and get explicit confirmation before retrying, since a blind retry on a partially-applied change can leave state inconsistent.
