TypeScript-based MCP (Model Context Protocol) server that connects Firefly with Claude, Cursor, etc., enabling resource discovery and IaC generation across your Cloud/SaaS accounts.
https://github.com/gofireflyio/firefly-mcpStop switching between AWS consoles, CLI commands, and Terraform docs. The Firefly MCP server connects your existing Firefly cloud inventory directly to Claude and Cursor, letting you discover resources and generate Infrastructure as Code using plain English.
You know the drill: find a resource in the AWS console, copy its configuration, open your Terraform files, figure out the right resource syntax, then manually translate everything. It's tedious, especially when you're working across multiple AWS accounts or need to codify existing resources that weren't originally managed as code.
This MCP server eliminates that friction. Your AI assistant now has direct access to your actual cloud inventory through Firefly, so you can go from discovery to working Terraform in a single conversation.
Instant Resource Discovery: Query any resource across all your connected cloud accounts without leaving your editor. No more tab-hopping between consoles or remembering account-specific CLI profiles.
Zero-Effort IaC Generation: Convert discovered resources directly into Terraform configuration. The AI understands your actual resource properties and generates syntactically correct code based on the real state.
Natural Language Queries: Skip complex AWS CLI syntax. Just describe what you're looking for: "Find all production RDS instances in the US-East region" or "Show me the load balancers handling our API traffic."
Audit and Codify Existing Infrastructure:
Find all t3.large EC2 instances in account 123456789012 that don't have backup tags and generate Terraform to manage them
Cross-Account Resource Discovery:
Show me all S3 buckets containing "analytics" across all connected accounts and create Terraform modules for each environment
Compliance and Cleanup Tasks:
Find RDS instances without encryption enabled and generate Terraform to recreate them with proper security settings
Migration Planning:
List all resources in our staging VPC and create Terraform configuration to replicate the setup in a new region
Works with your existing setup - no architectural changes needed. If you're already using Firefly for cloud asset management and Claude/Cursor for development, this just connects the two.
Set up takes minutes:
# Quick start
npx @fireflyai/firefly-mcp --access-key your_key --secret-key your_secret
# Or configure in your mcp.json
{
"mcpServers": {
"firefly": {
"command": "npx",
"args": ["-y", "@fireflyai/firefly-mcp"],
"env": {
"FIREFLY_ACCESS_KEY": "your_access_key",
"FIREFLY_SECRET_KEY": "your_secret_key"
}
}
}
}
This isn't about replacing your existing tools - it's about connecting them more intelligently. Your cloud resources become queryable context for your AI coding assistant, turning infrastructure discovery and codification from a multi-step process into a conversation.