# Does OpenSearch have an MCP server?

**Yes — Official.** OpenSearch maintains an official MCP server.

## Facts

- Verdict: Official (Yes)
- Access: Built into the product — Ships inside the vendor's own app or CLI rather than as a standalone download.
- Endpoint: `Built into OpenSearch 3.3+ at /_plugins/_ml/mcp (ML Commons, in-product)`
- Package: [`opensearch-mcp-server-py`](https://pypi.org/project/opensearch-mcp-server-py/) on pypi
- Repository: [opensearch-project/opensearch-mcp-server-py](https://github.com/opensearch-project/opensearch-mcp-server-py) — ★ 155, last push 2026-09-19, Apache-2.0
- Maintainer: OpenSearch Project
- Language: Python
- Category: [Databases](https://mcpyet.com/category/databases/)
- Verified against vendor docs: 2026-09-29
- GitHub data refreshed: 2026-09-29 (re-checked every 6 hours)

## The receipts

- **In-product:** Built into OpenSearch 3.3+ at /_plugins/_ml/mcp (ML Commons, in-product) — ships inside OpenSearch's own product rather than as a standalone server. Documented at <https://docs.opensearch.org/latest/ml-commons-plugin/api/mcp-server-apis/mcp-server/>.
- **Official package:** [`opensearch-mcp-server-py`](https://pypi.org/project/opensearch-mcp-server-py/) on pypi. The package page carries the exact launch arguments and required credentials.
- **Official repository:** [opensearch-project/opensearch-mcp-server-py](https://github.com/opensearch-project/opensearch-mcp-server-py) — ★ 155, pushed 2026-09-19, Apache-2.0 (active)

## How to connect it

Ships inside the vendor's own app or CLI rather than as a standalone download.

### Claude Code

```sh
claude mcp add opensearch -- npx -y <package-from-readme>
```

Swap in the exact package and flags from the repo README — most servers need an API key passed as an environment variable. Add `--scope user` to make it available in every project.

### Claude Desktop

Open **Settings → Developer → Edit Config** and add the server to `claude_desktop_config.json`, then restart Claude Desktop:

```json
{
  "mcpServers": {
    "opensearch": {
      "command": "npx",
      "args": [
        "-y",
        "<package-from-readme>"
      ],
      "env": {
        "API_KEY": "…"
      }
    }
  }
}
```

### Cursor

Add to `.cursor/mcp.json` in your project, or `~/.cursor/mcp.json` for every project:

```json
{
  "mcpServers": {
    "opensearch": {
      "command": "npx",
      "args": [
        "-y",
        "<package-from-readme>"
      ],
      "env": {
        "API_KEY": "…"
      }
    }
  }
}
```

### VS Code

Add to `.vscode/mcp.json`, then pick the server from Copilot's agent-mode tools menu:

```json
{
  "servers": {
    "opensearch": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "<package-from-readme>"
      ]
    }
  }
}
```

### ChatGPT

ChatGPT connects to **remote servers only**, and this one runs on your machine. Either expose it through an MCP gateway that gives it a public URL, or use a client with native local support (Claude or Cursor).

## How we know

docs.opensearch.org documents the OpenSearch MCP Server, built from opensearch-project/opensearch-mcp-server-py in the verified opensearch-project org and installed from PyPI as opensearch-mcp-server-py. Since 3.3, the ML Commons plugin also exposes a built-in Streamable HTTP MCP endpoint at /_plugins/_ml/mcp.

- GitHub data refreshed 2026-09-29, then every 6 hours.
- Vendor docs verified on 2026-09-29.
- Corrections: hello@mcpyet.com.

## Also in Databases

- [Snowflake](https://mcpyet.com/mcp/snowflake.md) — Official (Yes)
- [Firebase](https://mcpyet.com/mcp/firebase.md) — Official (Yes)
- [Supabase](https://mcpyet.com/mcp/supabase.md) — Official (Yes)
- [PostgreSQL](https://mcpyet.com/mcp/postgres.md) — Community only (Sort of)
- [BigQuery](https://mcpyet.com/mcp/bigquery.md) — Official (Yes)
- [ClickHouse](https://mcpyet.com/mcp/clickhouse.md) — Official (Yes)

---

Source: https://mcpyet.com/mcp/opensearch/ — data refreshed 2026-09-29