# Valca MCP server

> Scans code, IaC, and AI agent configurations to block insecure file writes during the generation process.

- Repository: https://github.com/PranLabs/valca
- Page: https://hostedmcp.org/servers/pranlabs/valca
- Transport: stdio
- Stars: 2
- Latest version: 0.6.1

## Install

### Claude Code

```sh
claude mcp add valca -- uvx valca
```

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

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

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

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

### VS Code CLI

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

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

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

### Codex

```sh
codex mcp add valca -- uvx valca
```

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

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

### Gemini CLI

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

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

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

## Environment variables

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

## FAQ

### Does Valca send my code to the cloud?

No, the core scanning is fully offline and the engine never sends your code, file paths, or findings anywhere.

### How does Valca interact with Claude Code?

Valca intercepts files written by AI assistants and can block the write if critical or high security findings are detected.

### Can I run Valca completely offline?

Yes, you can disable specific network rules in your .valcarc file to ensure no package information is sent to external registries.

### What kind of security issues does it catch?

It detects a wide range of issues including hardcoded secrets, insecure Docker configurations, Terraform vulnerabilities, and prompt injection risks.
