# PrismSearch MCP server

> Aggregates web, image, and WeChat search results from multiple engines and fetches page or PDF content for AI agents.

- Repository: https://github.com/xiaonerner0821/prismsearch
- Page: https://hostedmcp.org/servers/xiaonerner0821/prismsearch
- Transport: stdio
- Stars: 0
- License: Apache-2.0
- Latest version: 0.3.3

## Install

### Claude Code

```sh
claude mcp add --env 'BRAVE_API_KEY=<YOUR_BRAVE_API_KEY>' --env 'GOOGLE_API_KEY=<YOUR_GOOGLE_API_KEY>' prismsearch -- npx -y prismsearch
```

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

```json
{
  "mcpServers": {
    "prismsearch": {
      "command": "npx",
      "args": [
        "-y",
        "prismsearch"
      ],
      "env": {
        "BRAVE_API_KEY": "<YOUR_BRAVE_API_KEY>",
        "GOOGLE_API_KEY": "<YOUR_GOOGLE_API_KEY>"
      }
    }
  }
}
```

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

```json
{
  "servers": {
    "prismsearch": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "prismsearch"
      ],
      "env": {
        "BRAVE_API_KEY": "<YOUR_BRAVE_API_KEY>",
        "GOOGLE_API_KEY": "<YOUR_GOOGLE_API_KEY>"
      }
    }
  }
}
```

### VS Code CLI

```sh
code --add-mcp '{"name":"prismsearch","type":"stdio","command":"npx","args":["-y","prismsearch"],"env":{"BRAVE_API_KEY":"<YOUR_BRAVE_API_KEY>","GOOGLE_API_KEY":"<YOUR_GOOGLE_API_KEY>"}}'
```

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

```json
{
  "mcpServers": {
    "prismsearch": {
      "command": "npx",
      "args": [
        "-y",
        "prismsearch"
      ],
      "env": {
        "BRAVE_API_KEY": "<YOUR_BRAVE_API_KEY>",
        "GOOGLE_API_KEY": "<YOUR_GOOGLE_API_KEY>"
      }
    }
  }
}
```

### Codex

```sh
codex mcp add prismsearch --env 'BRAVE_API_KEY=<YOUR_BRAVE_API_KEY>' --env 'GOOGLE_API_KEY=<YOUR_GOOGLE_API_KEY>' -- npx -y prismsearch
```

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

```toml
[mcp_servers.prismsearch]
command = "npx"
args = ["-y", "prismsearch"]

[mcp_servers.prismsearch.env]
BRAVE_API_KEY = "<YOUR_BRAVE_API_KEY>"
GOOGLE_API_KEY = "<YOUR_GOOGLE_API_KEY>"
```

### Gemini CLI

```sh
gemini mcp add -s user -e 'BRAVE_API_KEY=<YOUR_BRAVE_API_KEY>' -e 'GOOGLE_API_KEY=<YOUR_GOOGLE_API_KEY>' prismsearch npx -- -y prismsearch
```

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

```json
{
  "mcpServers": {
    "prismsearch": {
      "command": "npx",
      "args": [
        "-y",
        "prismsearch"
      ],
      "env": {
        "BRAVE_API_KEY": "<YOUR_BRAVE_API_KEY>",
        "GOOGLE_API_KEY": "<YOUR_GOOGLE_API_KEY>"
      }
    }
  }
}
```

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `BRAVE_API_KEY` | no | yes | Brave Search API key (optional, comma-separated for rotation) |
| `GOOGLE_API_KEY` | no | yes | Google Custom Search API key (optional, comma-separated for rotation) |
| `GOOGLE_CSE_ID` | no | no | Google Programmable Search Engine ID (cx), used with GOOGLE_API_KEY |
| `PRISMSEARCH_AUTH_TOKENS` | no | no |  |
| `PRISMSEARCH_CACHE_BACKEND` | no | no |  |
| `PRISMSEARCH_CACHE__BACKEND` | no | no |  |
| `PRISMSEARCH_CACHE__TTL` | no | no |  |
| `PRISMSEARCH_CONFIG` | no | no | Path to a YAML/JSON/TOML config file (optional) |
| `PRISMSEARCH_ENGINES__BING__ENABLED` | no | no |  |
| `PRISMSEARCH_ENGINES__DUCKDUCKGO__ENABLED` | no | no |  |
| `PRISMSEARCH_FETCHER__BROWSER__ENABLED` | no | no |  |
| `PRISMSEARCH_HOME` | no | no | Local data directory for SQLite cache and default config lookup (optional) |
| `PRISMSEARCH_REDIS_PASSWORD` | no | no |  |
| `PRISMSEARCH_SECURITY__RATE_LIMIT__MAX_REQUESTS` | no | no |  |
| `PRISMSEARCH_SECURITY__SSRF__WHITELIST` | no | no |  |
| `PRISMSEARCH_SERVER_TRANSPORT` | no | no |  |
| `PRISMSEARCH_SERVER__ALLOWED_HOSTS` | no | no |  |
| `PRISMSEARCH_WORKERS_ENABLED` | no | no |  |
| `PRISMSEARCH_WORKER_TOKEN` | no | no |  |

## FAQ

### Can PrismSearch read PDF files?

Yes, the fetch_page tool can read web pages and PDFs, providing Markdown output.

### Does it filter out advertisements?

Yes, the server removes sponsored results from search engines and filters landing URLs to drop paid-click tracking parameters.
