# TerraVision MCP server

> Converts Terraform code into professional cloud architecture diagrams for AWS, Azure, and GCP using AI-powered refinement.

- Repository: https://github.com/patrickchugh/terravision
- Page: https://hostedmcp.org/servers/patrickchugh/terravision
- Transport: stdio
- Stars: 1690
- License: AGPL-3.0
- Latest version: 0.56.1

## Install

### Claude Code

```sh
claude mcp add terravision -- uvx terravision
```

### Cursor (`~/.cursor/mcp.json`)

```json
{
  "mcpServers": {
    "terravision": {
      "command": "uvx",
      "args": [
        "terravision"
      ]
    }
  }
}
```

### VS Code (`.vscode/mcp.json`)

```json
{
  "servers": {
    "terravision": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "terravision"
      ]
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"terravision","type":"stdio","command":"uvx","args":["terravision"]}'
```

### Claude Desktop (`claude_desktop_config.json`)

```json
{
  "mcpServers": {
    "terravision": {
      "command": "uvx",
      "args": [
        "terravision"
      ]
    }
  }
}
```

### Codex

```sh
codex mcp add terravision -- uvx terravision
```

### Codex config.toml (`~/.codex/config.toml`)

```toml
[mcp_servers.terravision]
command = "uvx"
args = ["terravision"]
```

### Gemini CLI

```sh
gemini mcp add -s user terravision uvx -- terravision
```

### Gemini settings.json (`~/.gemini/settings.json`)

```json
{
  "mcpServers": {
    "terravision": {
      "command": "uvx",
      "args": [
        "terravision"
      ]
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `OLLAMA_KEEP_ALIVE` | no | no |  |

## FAQ

### Does TerraVision require access to my cloud environment?

No, it runs 100% client-side without any dependency or access to your cloud environment, working off your terraform plan rather than your remote statefile.

### Can I use local AI models for diagram refinement?

Yes, you can use Ollama as a local AI backend for privacy-focused, offline AI refinement.

### Can this be integrated into a CI/CD pipeline?

Yes, it is designed as a 'Docs as Code' tool that can be included in your CI/CD pipeline to update diagrams automatically after build or test phases.