AI Integration (MCP)

Give your AI assistant (Cursor, Claude Desktop, Windsurf) full SmoothSend context — so it can generate correct integration code, estimate costs, and answer SDK questions without hallucinating.

@smoothsend/mcp on npm

What is MCP?

Model Context Protocol — an open standard from Anthropic

MCP is a protocol that lets AI assistants inside your editor connect to external servers that expose structured knowledge and callable functions. The server runs as a local subprocess on your machine, communicating with the editor's AI over stdin/stdout.

Once configured, when you ask your AI "how do I integrate SmoothSend?" it calls the right tool on the MCP server, gets back accurate up-to-date content, and generates working code — instead of guessing from training data.

Setup — 2 minutes

No install required — just add the config to your editor

Cursor

Add to your global or project-level Cursor config

~/.cursor/mcp.json

JSON

{
  "mcpServers": {
    "smoothsend": {
      "command": "npx",
      "args": ["@smoothsend/mcp"]
    }
  }
}

Claude Desktop

Add to your Claude Desktop config (macOS)

~/Library/Application Support/Claude/claude_desktop_config.json

JSON

{
  "mcpServers": {
    "smoothsend": {
      "command": "npx",
      "args": ["@smoothsend/mcp"]
    }
  }
}

Windsurf

Add to your Windsurf MCP config

~/.codeium/windsurf/mcp_config.json

JSON

{
  "mcpServers": {
    "smoothsend": {
      "command": "npx",
      "args": ["@smoothsend/mcp"]
    }
  }
}

Restart your editor after saving the config. The MCP server starts automatically — no separate process to run.

Available Tools

Functions the AI can call to get accurate SmoothSend information

get_docs

Fetch full documentation for any section — overview, installation, quickstart, api-reference, examples, or billing.

estimate_credits

Calculate monthly credit cost from gas_used, gas_unit_price, apt_price_usd, and transaction volume. Returns per-tx and monthly totals.

get_token_address

Get the Aptos Mainnet fungible asset address for USDC, USDT, WBTC, USDe, or USD1 with a ready-to-paste code snippet.

get_code_snippet

Return a complete, copy-pasteable TypeScript snippet for wallet-adapter-setup, script-composer-usdc, fee-preview, error-handling, or testnet-setup.

Resources

Documentation pages the AI can read directly as reference material

text

smoothsend://docs/overview        — Overview, quick start, pricing
smoothsend://docs/installation    — Installation guide
smoothsend://docs/quickstart      — Step-by-step quickstart
smoothsend://docs/api-reference   — Full API reference
smoothsend://docs/examples        — Real-world code examples
smoothsend://docs/billing         — Pricing and credit details

Example Prompts

Try these in your AI assistant once the MCP server is configured

›"How do I add SmoothSend to my Next.js app?"
›"Show me how to send USDC without the user needing APT."
›"My dApp does 5,000 contract calls per month at 200 gas each. How much will SmoothSend cost?"
›"What is the USDC asset address on Aptos Mainnet?"
›"How do I handle the error when SmoothSend credits run out?"
›"Give me a testnet setup that is completely free."

How it works

When you run npx @smoothsend/mcp, it starts a lightweight Node.js process that speaks the MCP JSON-RPC protocol over stdin/stdout. Your editor's AI connects to it at startup and discovers the available tools and resources.

All content is bundled directly in the npm package — no network calls, no API keys for the MCP itself. When the AI needs SmoothSend information, it calls a tool locally (sub-millisecond), gets back structured markdown, and uses it to generate accurate code for your project.

The package is fully open source on GitHub. To keep content up to date as the SDK evolves, the package is updated alongside each SDK release.