Azure Cost Calculator - AI Agent Plugin
Real-time Azure cost estimation using the public Azure Retail Prices API. Works with any agent in the skills.sh ecosystem. All prices come from live API lookups. No Azure subscription required.
Install
Install as a Plugin (recommended for Copilot CLI and Claude Code)
Run inside a Copilot CLI or Claude Code session.
/plugin marketplace add ahmadabdalla/plugins
/plugin install azure-cost-calculator@ahmadabdalla-plugins
Plugin install pulls from versioned releases with changelog tracking and update control. For the full plugin install/uninstall/update guide, see Plugin Management.
Install as a Skill (works with any agent)
npx (works with any agent):
npx skills add ahmadabdalla/azure-cost-calculator-skill
Don't have npx? Install Node.js (which includes npm and npx), or run npm install -g skills first then use skills add ahmadabdalla/azure-cost-calculator-skill.
The npx method pulls the latest from the main branch directly; it always gets the current stable content but without version pinning or rollback.
Usage
Ask about Azure costs in natural language. The skill activates automatically.
How much does a D4s v5 VM cost per month in East US?
Compare App Service pricing tiers for a production web app
Estimate a Standard_B2s VM with a P30 managed disk in Australia East in AUD
What's the cost of a General Purpose SQL Database with 4 vCores in West Europe in EUR?
How much would Azure Cosmos DB with 1000 RU/s and 100 GB storage cost?
Planning a larger architecture? Start with:
I'd like to perform a cost analysis on an Azure architecture. What do I need to consider to get consistent results?
The agent will walk you through the key parameters that affect pricing accuracy. For the full guide on writing prompts that produce deterministic estimates, see the Usage Guide.
How It Works
The skill uses service reference files as an index. Each file contains exact API filter values as declarative Key: Value parameters, cost formulas, and traps. The agent reads the matching file, translates the parameters to the detected runtime (Bash or PowerShell), runs the pricing script against the live API, and presents a structured estimate.
The skill optimises for two goals:
- Determinism - target ≤ 5% cost variance. Same query → same API call → same price. All values from the live API, nothing hardcoded. LLMs are non-deterministic by nature, so this skill is designed to constrain them where possible: pre-verified filters, explicit formulas, and scripted API calls reduce the surface area where the model can drift.
- Token efficiency - target ≤ 5% token usage variance. Only SKILL.md and shared.md load on every query. Service files load on demand. Batch mode (3+ services) reads only the first 45 lines per file.
Other design goals:
- Multi-currency, all regions - supports USD, AUD, EUR, GBP, JPY, CAD, INR, etc. Works with any Azure region.
Note: Targets measured via A/B testing with clean-context sessions against complex Azure architectures. Tested with Claude Opus 4.6 and Gemini Pro 3. Results with other models may vary.
References load on demand, keeping token usage low even for 10+ service estimates.
Supported Services
210+ Azure services are mapped across 18 categories (Compute, Databases, Networking, Storage, Security, Monitoring, Integration, AI + ML, and more). 110+ services have full reference files with documented query patterns. For services without a reference file, the skill includes an exploration script that searches the live API to find the right filters automatically.
Found a Gap? Open an Issue
If you query a service and the skill falls back to discovery mode, that's a signal we're missing a reference file. Please open an issue with the service name rather than accepting the best-effort result. Even if the estimate looked correct this time, the next user (or the next API change) may not get the same result. Issues help us prioritise which reference files to add next.
Prerequisites