# MCP Servers (mobile)
An MCP server, short for Model Context Protocol server, is a lightweight program that exposes specific capabilities to AI models via a standardized protocol, enabling AI to interact with external data sources and tools.
MCP servers allows you to extend BoltAI's capabilities with custom tools and commands.
{% hint style="warning" %}
This guide is for the mobile app. To set up MCP servers for the Mac app, [follow this guide](https://docs.boltai.com/docs/plugins/mcp-servers).
{% endhint %}
## Adding a server
You can only use a remote MCP server in BoltAI for iOS as the mobile OS doesn't allow running a local server. Note that you can add an stdio server when using the Mac app. Make sure your server is reachable from the your phone (via SSE). To use a server on your Mac, make sure your phone and your Mac are connected to the same network. For convenience, set your Mac to use a static IP instead of DHCP.
Adding a server is simple. Go to Settings > MCP Servers, tap Add. Enter the server name, and your remote MCP server URL. Tap "Add Server".
## Guides
### How to setup Zapier MCP server?
Zapier is an integration platform that hosts 7,000+ third-party apps and has 30,000+ actions. AI Actions equips your own AI or systems with the ability to run any action on Zapier (eg. Send a Slack message, find a Google Calendar event, draft a Gmail reply, create a HubSpot lead, etc.)
Using Zapier MCP server, you can build your own AI assistant to perform real-world tasks.
[Read the official guide](https://actions.zapier.com/settings/mcp/)
Getting Started with Zapier MCP server
### How to use a local MCP server?
To run a local MCP server (stdio transport), you will need to convert it to a SSE server using [supergateway](https://github.com/supercorp-ai/supergateway). If you need to access it from the Internet, either host the MCP server using Cloudflare or use a tunnel service like ngrok or Cloudflare Tunnel.
{% hint style="warning" %}
**IMPORTANT**: Without authentication, anyone with the URL can interact with your MCP server. Make sure to keep it secret and turn the server off when not in use.
{% endhint %}
For example, here is a step-by-step guide on how I set up the Memory MCP server and expose it to the internet via ngrok.
**Start the SSE server using supergateway:**
```bash
npx -y supergateway \
--stdio "npx -y @modelcontextprotocol/server-memory" \
--port 8071 \
--ssePath /sse \
--messagePath /message
```
**Serve it using ngrok (optional):**
```bash
$> ngrok http 8071
ngrok (Ctrl+C to quit)
👋 Goodbye tunnels, hello Agent Endpoints: https://ngrok.com/r/aep
Session Status online
Account Daniel Nguyen (Plan: Free)
Version 3.22.0
Region Asia Pacific (ap)
Latency 95ms
Web Interface http://127.0.0.1:4040
Forwarding https://xxxx-xxxx-xxx-xxxx-xxxx-xxxx-xxxx-xxxx-xxxx.ngrok-free.app -> http://localhost:8071
Connections ttl opn rt1 rt5 p50 p90
0 0 0.00 0.00 0.00 0.00
```
**Add to BoltAI:**
If you don't use ngrok, your server would look like this: `http://:8071/sse`
If you use ngrok, your remote MCP server URL would be `https://xxxx-xxxx-xxx-xxxx-xxxx-xxxx-xxxx-xxxx-xxxx.ngrok-free.app/sse`
Now you can use this MCP server on the mobile app ✌️
## Recommended MCP services:
Some interesting remote MCP services to try:
* [Zapier MCP](https://zapier.com/mcp): Connect your AI to any app with Zapier MCP
* [Composio](https://mcp.composio.dev): Instantly Connect to 100+ Managed MCP Servers with Built-In Auth
* [GitMCP](https://gitmcp.io): Instantly create a Remote MCP server for any GitHub project
* [MCPify](https://mcpify.ai): Create Your MCP Server using LLMs
