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) or gcloud auth application-default login
- Requires:
GCP_PROJECT_ID environment 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_projects to list all accessible GCP projects
- List instances:
list_instances per project — name, status, zone, machine type
- Network interfaces:
get_instance_basic_info to see IPs, VPC, subnet per VM
- Instance groups:
list_instance_group_managers for 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_instances filtered 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-logging for instance logs, gcp-cloud-monitoring for 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 use gcloud auth application-default login)
1---2name: gcp-compute-ops3description: Google Cloud Compute Engine — VM instances, disks, templates, instance groups, reservations, project discovery. Use when listing GCP VMs, troubleshooting a Compute Engine instance, checking disk attachments, planning capacity with reservations and commitments, or managing instance groups.4license: Apache-2.05---6
7# GCP Compute Operations
8
9## MCP Servers
10
11- **Compute Engine**: `https://compute.googleapis.com/mcp` (Streamable HTTP)
12- **Resource Manager**: `https://cloudresourcemanager.googleapis.com/mcp` (Streamable HTTP)
13- **Auth**: OAuth 2.0 via Google IAM — service account key (`GOOGLE_APPLICATION_CREDENTIALS`) or `gcloud auth application-default login`
14- **Requires**: `GCP_PROJECT_ID` environment variable
15
16## Key Difference from On-Prem MCP Servers
17
18Google 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.
19
20## Available Tools (29)
21
22### Resource Manager (1 tool)
23
24| Tool | What It Does |
25|------|-------------|
26| `search_projects` | Search for GCP projects by parent, ID, or filter. Use this first to discover available projects. |
27
28### Instance Management (8 tools)
29
30| Tool | What It Does |
31|------|-------------|
32| `list_instances` | List VM instances in a project (filter by zone) |
33| `get_instance_basic_info` | Get core instance details — name, status, machine type, network interfaces, IPs |
34| `create_instance` | Provision a new VM instance |
35| `delete_instance` | Remove a VM instance |
36| `start_instance` | Boot a stopped instance |
37| `stop_instance` | Halt a running instance |
38| `reset_instance` | Restart an instance (hard reset) |
39| `set_instance_machine_type` | Resize an instance (change machine type) |
40
41### Instance Groups (3 tools)
42
43| Tool | What It Does |
44|------|-------------|
45| `list_instance_group_managers` | List managed instance groups (MIGs) |
46| `get_instance_group_manager_basic_info` | Get MIG details — template, target size, autoscaling |
47| `list_managed_instances` | List instances within a MIG |
48
49### Instance Templates (3 tools)
50
51| Tool | What It Does |
52|------|-------------|
53| `list_instance_templates` | List available instance templates |
54| `get_instance_template_basic_info` | Get template basics — name, description |
55| `get_instance_template_properties` | Get template specs — machine type, disks, network interfaces |
56
57### Disks & Storage (5 tools)
58
59| Tool | What It Does |
60|------|-------------|
61| `list_disks` | List persistent disks in a project |
62| `get_disk_basic_info` | Get disk metadata — size, type, attached instances |
63| `get_disk_performance_config` | Get disk performance settings — IOPS, throughput |
64| `list_snapshots` | List disk snapshots |
65| `list_instance_attached_disks` | List disks attached to a specific instance |
66
67### Resource Discovery (3 tools)
68
69| Tool | What It Does |
70|------|-------------|
71| `list_machine_types` | List available machine types in a zone (e2, n2, c3, etc.) |
72| `list_accelerator_types` | List available GPU/TPU types in a zone |
73| `list_images` | List available boot images (Debian, Ubuntu, CentOS, etc.) |
74
75### Reservations & Commitments (6 tools)
76
77| Tool | What It Does |
78|------|-------------|
79| `list_reservations` | List capacity reservations |
80| `get_reservation_basic_info` | Get reservation details |
81| `get_reservation_details` | Get comprehensive reservation info |
82| `list_commitments` | List committed use discounts (CUDs) |
83| `get_commitment_basic_info` | Get commitment details — term, status |
84| `list_commitment_reservations` | List reservations tied to a commitment |
85
86### Operations (1 tool)
87
88| Tool | What It Does |
89|------|-------------|
90| `get_zone_operation` | Check status of an async operation (create, delete, resize) |
91
92## Workflow: GCP Infrastructure Audit
93
94When a user asks "show me our GCP compute resources":
95
961. **Discover projects**: `search_projects` to list all accessible GCP projects
972. **List instances**: `list_instances` per project — name, status, zone, machine type
983. **Network interfaces**: `get_instance_basic_info` to see IPs, VPC, subnet per VM
994. **Instance groups**: `list_instance_group_managers` for autoscaled workloads
1005. **Disk inventory**: `list_disks` — size, type, attached instances
1016. **Report**: GCP compute inventory with status and network details
102
103## Workflow: VM Troubleshooting
104
105When investigating a GCP VM issue:
106
1071. **Find the VM**: `list_instances` filtered by name or zone
1082. **Get details**: `get_instance_basic_info` — status, machine type, network interfaces, external IP
1093. **Check disks**: `list_instance_attached_disks` — any missing or detached?
1104. **Check operation**: `get_zone_operation` — any pending/failed operations?
1115. **Cross-reference**: Use `gcp-cloud-logging` for instance logs, `gcp-cloud-monitoring` for metrics
1126. **Report**: VM health assessment with findings
113
114## Workflow: Capacity Planning
115
116When planning GCP compute capacity:
117
1181. **Current usage**: `list_instances` — count by machine type and zone
1192. **Reservations**: `list_reservations` — reserved capacity vs actual usage
1203. **Commitments**: `list_commitments` — CUD utilization and expiration
1214. **Available types**: `list_machine_types` — what's available in target zones
1225. **GPU availability**: `list_accelerator_types` — GPU/TPU options per zone
1236. **Report**: Capacity utilization with recommendations
124
125## Important Rules
126
127- **Google Cloud MCP servers are remote** — no local install, runs on Google's infrastructure
128- **OAuth 2.0 authentication** — uses service account or user credentials via IAM
129- **Project-scoped** — all operations are scoped to the configured GCP project
130- **Write operations available** — unlike AWS Network MCP, Compute Engine MCP can create/modify/delete VMs. Use ServiceNow CR gating for changes.
131- **Record in GAIT** — log all GCP operations for audit trail
132
133## Environment Variables
134
135- `GCP_PROJECT_ID` — Google Cloud project ID
136- `GOOGLE_APPLICATION_CREDENTIALS` — Path to service account key JSON file (or use `gcloud auth application-default login`)
gabrielmoreira