GCP Cost Optimization MCP Server

🚀 Installation

Prerequisites

• The uv Python package and project manager from Astral or the GitHub README
• A GCP account
• Configured GCP credentials
• Install gcloud CLI for authentication

GCP Client Requirements

1. Credentials: Configure GCP credentials via gcloud CLI or environment variables
2. Permissions: Ensure your GCP credentials have the required permissions (see Permissions)

• Optional: GitHub token for PR automation, Slack webhook for notifications

Local development

To make changes to this MCP locally and run it:

# Clone this repository
git clone https://github.com/rohithay/dataops-mcp-server.git
cd dataops-mcp-server

# Create a virtual environment and install dependencies
uv sync
source .venv/bin/activate # On Windows, use `.venv\Scripts\activate`

🚦 Quick Start

1. Make sure to have configured your GCP credentials as described here
2. Update your claude_desktop_config.json file with proper configuration outlined in the AI integration guide
3. Open Claude for Desktop and start prompting! For more examples and advanced usage, see the detailed usage guide.

🤖 AI Integration

To get started using this MCP server with your AI assistants offering MCP support, like Claude Desktop, Cursor. Refer to this Integration guide.

Available Toolsets

The following sets of tools are available (all are on by deafult):

Toolset	Description
get_bigquery_costs	Retrieve BigQuery cost breakdowns
analyze_query_cost	Predict query cost before execution
detect_cost_anomalies	Detect unusual spending using ML
optimize_query	Use LLMs to auto-optimize SQL queries
create_optimization_pr	Auto-generate GitHub PRs
send_cost_alert	Send cost alerts to Slack
get_dbt_model_costs	Analyze dbt model costs
monitor_sla_compliance	Track SLA compliance and cost-performance trade-offs
forecast_costs	Forecast future costs
slack_post_message	Post message to Slack

---

Tools

Actions

Tool: get_bigquery_costs

Retrieve comprehensive BigQuery cost analysis for specified time periods.

Parameters	Type	Description
days	number (default: 7)	Number of days to analyze (1–90)
project_id	string (optional)	Specific GCP project ID
include_predictions	boolean (default: true)	Include ML-based cost forecasting
group_by	array (default: ["date"])	Grouping dimensions (date, user, dataset, query_type)
include_query_details	boolean (default: false)	Include individual query cost breakdowns

Actions

Tool: analyze_query_cost

Predict cost of a SQL query before execution and get optimization suggestions.

Parameters	Type	Description
sql	string (required)	SQL query to analyze
project_id	string (optional)	GCP project ID
include_optimization	boolean (default: true)	Include AI-powered optimization suggestions
optimization_model	string (default: "claude")	Model to use ("claude", "gpt-4")
create_pr_if_savings	boolean (default: false)	Create GitHub PR if savings exceed threshold

Actions

Tool: detect_cost_anomalies

Use ML to detect cost spikes, anomalies, and early signs of overruns.

Parameters	Type	Description
days	number (default: 30)	Historical days to analyze
sensitivity	string (default: "medium")	Sensitivity (low, medium, high)
project_id	string (optional)	GCP project ID
alert_threshold	number (default: 0.25)	Alert threshold (e.g., 0.25 = 25% increase)
send_slack_alert	boolean (default: false)	Send alert to Slack

Actions

Tool: optimize_query

LLM-powered query optimization with cost-saving recommendations.

Parameters	Type	Description
sql	string (required)	SQL query to optimize
optimization_goals	array (default: ["cost", "performance"])	Objectives for optimization
preserve_results	boolean (default: true)	Ensure results are unchanged
include_explanation	boolean (default: true)	Include explanation of changes
target_savings_pct	number (default: 30)	Target savings percentage
dbt_model_path	string (optional)	Path to dbt model for context

Actions

Tool: create_optimization_pr

Auto-create GitHub PRs with optimized SQL and validation tests.

Parameters	Type	Description
optimization_id	string (required)	Optimization analysis ID
repository	string (default: "data-platform")	GitHub repo name
base_branch	string (default: "main")	Base branch for the PR
title_prefix	string (default: "🚀 Cost Optimization")	Prefix for PR title
assign_reviewers	boolean (default: true)	Auto-assign reviewers
include_tests	boolean (default: true)	Generate validation tests

Actions

Tool: send_cost_alert

Send actionable cost alerts to Slack with rich context.

Parameters	Type	Description
alert_type	string (required)	Type of alert (anomaly, budget_warning, optimization_opportunity)
cost_data	object (required)	Data to include in the alert
severity	string (default: "medium")	Alert severity level
channel	string (default: "#data-ops-alerts")	Slack channel for alert
mention_users	array (default: [])	Users to mention
include_remediation	boolean (default: true)	Include fix suggestions

Actions

Tool: get_dbt_model_costs

Analyze dbt model execution costs and optimization opportunities.

Parameters	Type	Description
model_path	string (optional)	Specific dbt model path
include_dependencies	boolean (default: true)	Analyze downstream impacts
materialization_analysis	boolean (default: true)	Suggest materialization strategy improvements
days	number (default: 7)	Time period for analysis
suggest_optimizations	boolean (default: true)	Include cost-saving suggestions

Actions

Tool: monitor_sla_compliance

Monitor pipeline SLAs and correlate with cost-performance metrics.

Parameters	Type	Description
sla_type	string (default: "all")	SLA type (latency, freshness, success_rate, all)
time_window	string (default: "24h")	Time window for analysis
include_cost_correlation	boolean (default: true)	Link SLA with cost data
alert_on_breach	boolean (default: false)	Send alerts for SLA breaches
optimization_suggestions	boolean (default: true)	Suggest cost-aware fixes

Actions

Tool: forecast_costs

Forecast future GCP spend using ML and scenario modeling.

Parameters	Type	Description
forecast_days	number (default: 30)	Days to forecast
include_confidence_intervals	boolean (default: true)	Include prediction ranges
breakdown_by	array (default: ["service"])	Forecast by (e.g., service, project)
scenario_analysis	boolean (default: false)	Include optimistic/pessimistic forecasts
budget_recommendations	boolean (default: true)	Suggest budget allocations

Actions

Tool: slack_post_message

Post a message to a Slack channel.

Parameters	Type	Description
channel_id	string	ID of the Slack channel
text	string	Message text to post