# ragstack MCP server

> Lets an AI client search, chat with, and scrape a serverless RAG knowledge base containing documents, images, video, and audio on AWS.

- Repository: https://github.com/HatmanStack/RAGStack-Lambda
- Page: https://hostedmcp.org/servers/hatmanstack/ragstack-lambda
- Transport: stdio
- Stars: 26
- License: Apache-2.0
- Latest version: 0.1.4

## Install

### Claude Code

```sh
claude mcp add ragstack-lambda -- uvx ragstack-mcp
```

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

```json
{
  "mcpServers": {
    "ragstack-lambda": {
      "command": "uvx",
      "args": [
        "ragstack-mcp"
      ]
    }
  }
}
```

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

```json
{
  "servers": {
    "ragstack-lambda": {
      "type": "stdio",
      "command": "uvx",
      "args": [
        "ragstack-mcp"
      ]
    }
  }
}
```

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

```json
{
  "mcpServers": {
    "ragstack-lambda": {
      "command": "uvx",
      "args": [
        "ragstack-mcp"
      ]
    }
  }
}
```

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

```toml
[mcp_servers.ragstack-lambda]
command = "uvx"
args = ["ragstack-mcp"]
```

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

```json
{
  "mcpServers": {
    "ragstack-lambda": {
      "command": "uvx",
      "args": [
        "ragstack-mcp"
      ]
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `RAGSTACK_API_KEY` | no | no | Your RAGStack API key (from Dashboard → Settings) |
| `RAGSTACK_GRAPHQL_ENDPOINT` | no | no | Your RAGStack GraphQL API URL (from Dashboard → Settings) |

## FAQ

### What file formats are supported for document processing?

The server supports text formats like HTML, CSV, JSON, and DOCX, as well as OCR for PDFs and images, and media files like MP4 and MP3.

### How much does it cost to run?

The architecture is serverless with scale-to-zero, costing approximately $7-10 per month for 1000 documents using Textract and Haiku.