cloudwright-ai-mcp 1.8.0

MCP server for Cloudwright architecture intelligence

cloudwright-ai-mcp

mcp-name: io.github.xmpuspus/cloudwright

MCP (Model Context Protocol) server for Cloudwright architecture intelligence.

Exposes Cloudwright's design, cost, validate, and export capabilities as MCP tools for use with Claude Code and other MCP-compatible clients.

Installation

pip install cloudwright-ai-mcp

Usage

cloudwright mcp

Or add to your MCP client configuration:

{
  "mcpServers": {
    "cloudwright": {
      "command": "cloudwright",
      "args": ["mcp"]
    }
  }
}

Available Tools

• design — Generate cloud architecture from natural language
• cost — Estimate monthly infrastructure costs
• validate — Check compliance against HIPAA, PCI-DSS, SOC 2, FedRAMP, GDPR
• export — Export to Terraform, CloudFormation, Mermaid, D2, C4, SBOM
• compare — Compare architectures across cloud providers
• review_architecture: free, offline scorer + linter + validator critique (no LLM)
• scan_compliance_controls: maps findings to framework control IDs; optional OSCAL 1.1.2 output
• plan_infrastructure: proves exported IaC is deployable (terraform validate/plan, read-only)
• chat_create_session — Create a persistent multi-turn design session
• chat_send — Send a message to an existing session
• chat_list_sessions — List all saved sessions
• chat_delete_session — Delete a session

Sessions persist to ~/.cloudwright/sessions/ and survive process restarts. Sessions older than CLOUDWRIGHT_MCP_SESSION_TTL_DAYS (default 7) are swept on server start and on every chat_list_sessions call.

License

MIT