GCP MCP
# GCP MCP Application
## Claude Desktop Integration
To enable GCP management capabilities in Claude Desktop, simply add the following configuration to your Claude Desktop MCP configuration:
```json
{
"gcp-mcp": {
"command": "uvx",
"args": [
"gcp-mcp"
]
}
}
```
That's it! No additional setup or credential manipulation is required. When you first ask Claude to interact with your GCP resources, a browser window will automatically open for you to authenticate and grant access. Once you approve the access, Claude will be able to manage your GCP resources through natural language commands.
Here are some example requests you can make:
Basic Operations:
- "Could you list my GCP projects?"
- "Show me my compute instances"
- "What storage buckets do I have?"
Resource Creation:
- "Please create a compute instance with 2GB RAM and 10GB storage, name it MCP-engine"
- "Create a new storage bucket called my-backup-bucket in us-central1"
- "Set up a new VPC network named prod-network with subnet 10.0.0.0/24"
Resource Management:
- "Stop all compute instances in the dev project"
- "Show me all instances that have been running for more than 24 hours"
- "What's the current CPU usage of my instance named backend-server?"
- "Create a snapshot of my database disk"
Monitoring and Alerts:
- "Set up an alert for when CPU usage goes above 80%"
- "Show me all critical alerts from the last 24 hours"
- "What's the current status of my GKE clusters?"
## Features
The application provides comprehensive coverage of GCP services:
### Resource Management
- Projects and quotas management
- Asset inventory
- IAM permissions
### Compute & Infrastructure
- Compute Engine instances
- Storage buckets and disks
- VPC networks and firewall rules
- Kubernetes Engine (GKE) clusters
### Databases & Storage
- Cloud SQL instances
- Firestore databases
- Cloud Storage
- Database backups
### Monitoring & Billing
- Metrics and alerts
- Billing information
- Uptime monitoring
- Resource usage tracking
### Coming Soon
- Deployment manager and infrastructure as code
## Installation
```bash
pip install gcp-mcp
```
## License
[MIT License](LICENSE)
Your contributions and issues are welcome !
TDQS
Scored across 56 tools
Most tools have distinct purposes targeting specific GCP resources and actions, with clear boundaries between services like Compute Engine, Cloud Storage, IAM, and monitoring. However, some overlap exists between list_assets and other listing tools, and say_hello is an outlier that could cause confusion.
Tool names follow a highly consistent verb_noun pattern throughout, such as create_instance, list_firewall_rules, get_bucket_details, and delete_object. The naming convention is uniform across all 56 tools, making them predictable and easy to understand.
With 56 tools, the count is excessive for a single server, making it overwhelming and difficult for agents to navigate efficiently. While GCP is broad, this many tools suggests poor scoping and could lead to confusion or performance issues in tool selection.
The toolset provides comprehensive coverage across multiple GCP services, including authentication, compute, storage, networking, IAM, and monitoring, with CRUD operations for many resources. Minor gaps exist, such as missing update operations for some resources like firewall rules or service accounts, but core workflows are well-supported.