DoiT MCP Skills

AI agent skills that bring DoiT Cloud Intelligence into your favorite AI clients — Claude Code, ChatGPT, Cursor, Codex, and Gemini CLI.

Query cloud costs, investigate anomalies, generate reports, and produce API integration code across AWS, GCP, Azure, Snowflake, Datadog, and more — all through natural language.

Skills

Skill	Description
doit-mcp-setup	Configure DoiT MCP in any supported AI client with correct credentials and OAuth
doit-mcp-reporting	Generate cost intelligence reports and run ad-hoc Cloud Analytics queries
doit-mcp-anomaly-investigation	Root-cause cost anomalies by correlating DoiT alerts with cloud-specific inspection
doit-mcp-api	Generate production-ready API call code (TypeScript, Python, Go, Bash) for DoiT Platform

Quick Start

Claude Code (Plugin)

Install as a plugin for the full experience — skills, MCP server, and tools in one package:

claude plugin marketplace add https://github.com/doitintl/doit-mcp-skills
claude plugin install doit-mcp@doit-mcp-skills

Then optionally run the setup skill to configure credentials:

/doit-mcp-setup

Claude Code (Skills Only)

If you only want the skills without the plugin wrapper:

claude install-skill https://github.com/doitintl/doit-mcp-skills

Then optionally run the setup skill to configure credentials:

/doit-mcp-setup

Other AI Clients (Cursor, Codex, Gemini CLI, ChatGPT)

The skills in this repo are plain Markdown files — any AI that can read a file can use them. To set up DoiT MCP in a non-Claude client:

1. Point your agent at the skill file — paste the contents of the relevant SKILL.md into your agent's context, or reference the file directly if your client supports file attachments.
2. Run setup — tell your agent: "Follow the instructions in this skill to configure DoiT MCP for [your client]."
3. The agent will write the correct config file and guide you through any OAuth steps.

Alternatively, ask your agent to run the doit-mcp-setup skill for your client:

• Cursor — per-project .cursor/mcp.json configuration
• ChatGPT — manual setup via Actions UI (requires developer mode + API key)
• Codex — ~/.codex/config.toml configuration
• Gemini CLI — ~/.gemini/settings.json configuration

DCI CLI (Optional)

The setup skill can also install the DoiT Cloud Intelligence CLI — a terminal companion that talks to the same DoiT API as the MCP server. Run the setup skill and choose the CLI option, or install it directly:

Platform	Command
macOS	brew install doitintl/dci-cli/dci
Windows (WinGet)	winget install DoiT.dci
Windows (Scoop)	scoop bucket add doitintl https://github.com/doitintl/dci-cli && scoop install dci
Linux (.deb / .rpm)	Download from Releases

After install, run dci login to authenticate and dci list-reports to verify.

What You Can Do

Cost Reporting

"Show me my top 5 cost drivers for the last 30 days" "Break down GCP costs by project for Q1"

Anomaly Investigation

"Show me recent cost anomalies" "Investigate the EC2 cost spike from last Tuesday"

API Code Generation

"Generate TypeScript code to query costs by service for the last month" "Show me a Python script that lists all my cloud reports"

Architecture

doit-mcp-skills/
├── .claude-plugin/          # Claude Code plugin metadata
│   ├── plugin.json          # Plugin manifest (skills + MCP config)
│   └── marketplace.json     # Marketplace manifest for plugin install
├── .mcp.json                # MCP server configuration
└── skills/
    ├── doit-mcp-setup/      # Client setup & configuration
    ├── doit-mcp-reporting/  # Report generation & queries
    ├── doit-mcp-anomaly-investigation/
    └── doit-mcp-api/        # API code generation

Each skill contains:

• SKILL.md — skill instructions and workflow
• agents/openai.yaml — agent configuration and tool dependencies
• references/ — detailed reference docs (dimensions, metrics, field mappings)
• templates/ — config snippets or output templates
• examples/ — working code examples (where applicable)

How It Works

All skills connect to the DoiT MCP server at https://mcp.doit.com/sse — no local server installation required. Authentication is handled via OAuth (automated clients) or API key (manual clients like ChatGPT).

Supported platforms: AWS, GCP, Azure, Snowflake, Datadog, Databricks, OpenAI

Prerequisites

• A DoiT account with Billing Profiles Admin permission
• An AI client that supports MCP (Claude Code, Cursor, ChatGPT, Codex, or Gemini CLI)

License

MIT