CloudForge MCP Server lets Claude visualise cloud architecture diagrams, generate/import Terraform, and manage CloudForge diagrams, organisation members, and AI architecture advice from chat.
Visualise diagrams — render a saved diagram as an inline Mermaid flowchart (cloudforge_diagram_to_mermaid), get a structured architecture summary with resource counts and hub components (cloudforge_architecture_summary), or convert raw diagram JSON to Mermaid offline (cloudforge_mermaid_from_json).
Generate Terraform — produce ready-to-use HCL from a natural-language description for Azure/AWS/GCP (cloudforge_generate_terraform) or straight from an existing saved diagram (cloudforge_export_terraform_from_diagram).
Import/understand existing IaC — parse raw Terraform HCL into diagram structure plus a Mermaid preview (cloudforge_import_terraform).
AI architecture help — ask the CloudForge AI Architect for IaC advice, cost, security or best-practice guidance (cloudforge_chat) and get AI-recommended resources and connections for a use case (cloudforge_suggest_resources).
Manage diagrams — list, fetch, save/update, delete, and quickly surface recently modified diagrams.
Manage the organisation — view members, invite people by email with a role, list and cancel pending invitations, and remove members.
(Per the README, the full server goes further with Terraform validation/repair, cost estimation, deployment plan/apply/destroy behind a human approval gate, drift and compliance scanning, self-healing remediations, governance policies, Molds, templates and pricing tools — but these are not present in the supplied schema.)
Converts CloudForge diagrams to Mermaid flowcharts for visualization, including tools to fetch diagrams and render them as inline Mermaid charts or convert raw diagram JSON to Mermaid format.
Provides comprehensive Terraform integration including generating HCL from natural-language descriptions, exporting HCL from existing diagrams, importing existing HCL to create diagram structures with Mermaid previews, and managing infrastructure as code.
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@CloudForge MCP Server generate Terraform for a simple AWS VPC with public and private subnets"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
MCP server for CloudForge — lets Claude and other AI assistants visualise cloud architecture diagrams, generate Terraform HCL, import existing IaC, and manage your infrastructure resources directly from chat.
Tool | Description |
| Fetch a saved diagram and render it as an inline Mermaid flowchart |
| Structured summary: resource counts by category, central components, inline diagram |
| Convert raw diagram JSON to Mermaid without fetching from the server |
Tool | Description |
| Generate Terraform HCL from a natural-language architecture description |
| Export Terraform HCL from an existing saved diagram in one step |
| Turn raw HCL into a diagram structure — deterministic, no AI, and no account needed |
| Export a saved diagram's Terraform as it stands, without regenerating |
| Parse existing HCL → diagram structure + Mermaid preview |
Tool | Description |
| Ask the CloudForge AI Architect for IaC advice, cost estimates, security reviews |
| Get AI-recommended resources and connections for a described use case |
Tool | Description |
| Browse 90 curated architecture templates (30 per cloud) — filter by provider, category, or search |
| Full detail for one Mold: layout, "why this shape" best practices, inline Mermaid preview, forge link |
Tool | Description |
| List saved diagrams with metadata |
| Get recently accessed/modified diagrams |
| Fetch a diagram by ID (full JSON) |
| Save or update a diagram |
| Permanently delete a diagram |
Tool | Description |
| Organisation overview: name, plan, member count, MFA policy, custom assistant name |
| List organisation members (name, email, role, status) |
| Send an organisation invitation |
| List pending invitations |
| Cancel a pending invitation |
| Remove a member from the organisation |
Tool | Description |
| Get CloudForge's current plans & pricing (Free, Pro, Team, Business), incl. the 14-day free trial — fetched live |
Real infrastructure changes. apply and destroy stop at a human approval gate and change
nothing until cloudforge_approve_deployment is called, so an agent cannot deploy on its own.
Tool | Description |
| Run |
| Validate and plan without touching a cloud account |
| Provider and HCL pre-flight checks |
| Start a real deployment. Creates resources and costs money; stops at the approval gate |
| Start a destroy. Permanently deletes infrastructure; same approval gate |
| Approve or reject a job waiting at its gate — the step that actually changes the cloud |
| Status and log of one job |
| Every deployment run against a diagram |
Tool | Description |
| Repair broken HCL from an error log — the same healer the deployment pipeline uses on itself |
| Syntax and provider-schema check without deploying |
| Estimated monthly cost, per resource |
| AI review: reliability, security, cost, scaling trade-offs |
Tool | Description |
| Compare a diagram against what is actually deployed |
| Most recent drift report for a diagram |
| Read one drift report by id — poll this after a scan |
| Frameworks available to scan against (APRA CPS 234, CIS, …) |
| Scan against frameworks — deterministic evidence and control mappings, never an AI opinion |
Three steps, always in that order. There is deliberately no tool that approves a fix: approval means a person read what would change and consented to it, and a tool that let an assistant grant that consent on their behalf would remove the only guarantee the feature offers. Detect and report freely; the user approves in the CloudForge panel, the VS Code extension or the JetBrains tool window; then heal.
Tool | Description |
| Turn a completed drift report into proposed fixes, all awaiting a person's decision |
| Proposals and where each one stands |
| Apply only what a person approved; pauses at the Terraform plan for a second yes |
| The append-only audit trail: what changed, why, and who said it could |
| The bounds healing works inside — excluded resource types, how long a proposal stays actionable |
Tool | Description |
| Governance policies configured for the organisation |
| Check a design against them before deploying — the pipeline enforces the same rules |
Tool | Description |
| Diagrams shared across the organisation (The Foundry) |
| Organisation-authored templates, distinct from the built-in Molds |
Related MCP server: Cloud Pilot MCP
Add to your claude_desktop_config.json:
{
"mcpServers": {
"cloudforge": {
"command": "npx",
"args": ["-y", "cloudforge-mcp"],
"env": {
"CLOUDFORGE_TOKEN": "your-bearer-token"
}
}
}
}.mcp.json{
"mcpServers": {
"cloudforge": {
"command": "npx",
"args": ["-y", "cloudforge-mcp"],
"env": {
"CLOUDFORGE_TOKEN": "your-bearer-token"
}
}
}
}Sign in at cloudforge.cloud
Go to Account → API Keys
Click Generate API Key, give it a name (e.g. CloudForge MCP), and copy the key — it starts with cf_live_ and is shown only once
Set it as CLOUDFORGE_TOKEN in your MCP config
The key is long-lived and can be revoked from the same page at any time.
Variable | Default | Description |
| (required) | CloudForge API key — must start with |
|
| API base URL — only change if self-hosting |
Show me my recent diagramsWhat serverless Molds do you have for AWS?Show me the azure-3-tier-web-app Mold and explain why it's shaped that wayRender diagram <id> as a Mermaid chartGenerate Terraform for a 3-tier Azure web app with a private SQL database in UK SouthImport this Terraform and show me a diagram: <paste HCL>What are the security best practices for the resources in diagram <id>?Invite john@example.com as a Member to my organisationThe cloudforge_get_pricing tool returns CloudForge's current plans live (so they're always up to date). All prices USD; paid plans include a 14-day free trial:
Plan | Price |
Free | $0 — visual designer, limited export, community support |
Pro | US$9.99/mo or US$99/yr · 14-day free trial |
Team | US$12/user/mo (3-seat min) · 14-day free trial |
Business | Contact us — volume/annual pricing |
Full details: cloudforge.cloud/pricing.
If you run CloudForge on-premises, point the server at your own API:
"env": {
"CLOUDFORGE_API_URL": "https://your-cloudforge-host/api",
"CLOUDFORGE_TOKEN": "your-token"
}MIT © CloudForge
Fetch a saved CloudForge diagram and return a structured summary: resource count by category, most-connected components, cloud provider, and an inline Mermaid diagram. Ideal for quickly understanding an architecture.
Inputs: diagram_id (required).
Cancel a pending organisation invitation by its ID.
Inputs: invitation_id (required).
Send a message to the CloudForge AI Architect. Use for IaC advice, best practices, cost optimisation, security reviews, or any cloud infrastructure question.
Inputs: message (required), provider.
Permanently delete a saved diagram by its ID.
Inputs: diagram_id (required).
Fetch a saved CloudForge diagram and render it as a Mermaid flowchart. The diagram renders visually inside Claude — this is the best way to SHOW a diagram. Resources are grouped by category (Compute, Network, Storage…) with labelled connections.
Inputs: diagram_id (required).
Fetch a saved CloudForge diagram and generate Terraform HCL for it in one step. Use when the user wants IaC code for an existing saved diagram.
Inputs: diagram_id (required), extra_instructions.
Generate Terraform HCL from a natural-language architecture description. Returns ready-to-use .tf files for Azure, AWS, or GCP.
Inputs: description (required), provider (required), diagram_json.
Fetch a saved diagram by ID. Returns full JSON: nodes, edges, resource configurations.
Inputs: diagram_id (required).
Parse existing Terraform HCL and return a diagram structure plus an inline Mermaid preview. Use when the user has .tf files and wants to visualise or understand their infrastructure.
Inputs: hcl (required).
Send an organisation invitation email via CloudForge.
Inputs: email (required), first_name (required), last_name (required), role.
List saved diagrams for the current user. Returns IDs, names, providers, and last-modified dates.
Inputs: limit.
List all pending organisation invitations (email, name, role, sent date, expiry).
Inputs: no named parameters.
List all members of the current user's CloudForge organisation. Returns name, email, role, and status.
Inputs: no named parameters.
Convert raw diagram JSON (nodes + edges) to a Mermaid flowchart without fetching from the server. Use this when you already have diagram data in memory.
Inputs: diagram_json (required).
Get recently accessed/modified diagrams (faster lookup than list_diagrams).
Inputs: no named parameters.
Remove a member from the organisation. Cannot remove the Owner. Pass the member's user ID from cloudforge_list_members.
Inputs: member_id (required).
Save or update a diagram in CloudForge.
Inputs: name (required), nodes (required), edges (required), cloud_provider, +1 more.
Get AI-recommended cloud resources for a described use case. Returns resource IDs, labels, categories, and suggested connections.
Inputs: description (required), provider (required).
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
Scored across 18 tools
Most tools have distinct purposes, but there is some overlap between cloudforge_diagram_to_mermaid and cloudforge_mermaid_from_json, which both generate Mermaid diagrams from diagram data. Similarly, cloudforge_export_terraform_from_diagram and cloudforge_generate_terraform both produce Terraform HCL, though from different inputs. Descriptions help clarify these differences, but slight confusion could occur.
All tool names follow a consistent snake_case pattern with the prefix 'cloudforge_' and a descriptive verb_noun structure (e.g., cloudforge_list_diagrams, cloudforge_save_diagram). This uniformity makes the tool set predictable and easy to navigate, with no deviations in naming conventions.
With 18 tools, the count is slightly high but reasonable for a cloud infrastructure diagramming and management server. It covers diagram operations, member management, and AI features, which justifies the breadth. However, it might feel a bit heavy compared to more focused servers, but each tool appears to serve a specific function.
The tool set provides comprehensive coverage for cloud infrastructure diagramming and management, including CRUD operations for diagrams (list, get, save, delete), member management (invite, list, remove), and advanced features like Terraform generation, Mermaid rendering, and AI assistance. No obvious gaps are present; it supports full lifecycle workflows from creation to analysis.
Unified API to query AWS, GCP, Azure and generate Terraform/CLI execution kits for AI agents.
Query your team's drift, vulnerability, and upgrade data from any AI assistant. OAuth 2.1, 51 tools.
Your AI Agent's Infrastructure Layer. Connect Claude, Copilot, Codex, or ChatGPT to 200+ managed open source services. Start databases, pipelines, and applications through natural language.
Plan Salesforce deploys, open pull requests and trigger pipelines from your AI client.
