> ## Documentation Index
> Fetch the complete documentation index at: https://docs.xentfi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Generic MCP Client

> Connect any other MCP-compatible client to XentFi agent tools.

If your client isn't covered by one of the specific guides, all it needs to know is how to run the stdio MCP server as a subprocess.

## Stdio configuration

| Setting            | Value                                                                                         |
| ------------------ | --------------------------------------------------------------------------------------------- |
| Command            | `npx`                                                                                         |
| Args               | `["-y", "@xentfi/mcp-server"]` (or `["-y", "@xentfi/mcp-server@<version>"]` to pin a version) |
| `XENTFI_AGENT_KEY` | Required — your agent's API key                                                               |
|                    |                                                                                               |

Once connected, the server advertises 15 tools, all prefixed `xentfi_` (see [Tools Reference](/agent/tools-reference)). It speaks standard MCP `tools/list` and `tools/call` — no client-specific behavior is required.

## If your client only supports remote/HTTP MCP servers

Use the `xentfi-mcp-http` binary instead (see [Remote HTTP Deployment](/agent/http-transport)) and point your client at:

* **URL**: `https://your-deployment/mcp`
* **Auth header**: `Authorization: Bearer <agent-api-key>` (plus `x-xentfi-org-id` if needed)

## Testing without any client at all

Use the official [MCP Inspector](https://github.com/modelcontextprotocol/inspector):

```bash theme={null}
XENTFI_AGENT_KEY=sk_agent_xxx npx @modelcontextprotocol/inspector npx -y @xentfi/mcp-server
```

This opens a browser UI where you can call every `xentfi_*` tool by hand and inspect raw request/response payloads — the fastest way to sanity-check a new API key or debug a tool before wiring it into an actual agent client.

## Next steps

<CardGroup cols={2}>
  <Card title="Tools Reference" icon="list" href="/agent/tools-reference">
    See every available tool.
  </Card>

  <Card title="Error Handling" icon="alert-triangle" href="/agent/errors">
    How failures surface to the calling agent.
  </Card>
</CardGroup>