🦞GMGN Agent API

1. Why choose GMGN AI Agent?

⚡️ Real-time On-chain Data · Provides multi-chain real-time market quotes and liquidity pool status, enabling your AI Agent to capture market opportunities with zero latency

🔍 Comprehensive Exclusive Data · The out-of-the-box API eliminates the need for tedious low-level data parsing. It directly outputs advanced research metrics, instantly identifying Snipers, Insiders, Bundled wallets, and more

🛡️ Secure and Convenient Trading · Adopts the GMGN hosted wallet architecture where private keys are not stored locally, eliminating leakage risks at the source. IP whitelist automatically intercepts abnormal trading requests, and developers no longer need to manually construct complex on-chain transactions


2. Run Your First GMGN AI Agent

Connect your AI Agent to on-chain capabilities—query quotes, fetch market data, and execute Swap transactions, all through natural language

Quick Start: Get Your API Key

Before using the tools, please create a GMGN API Key, 👉 Apply here: https://gmgn.ai/aiarrow-up-right


1. Add GMGN Skills to Your Agent

Copy and send the following prompt to your OpenClaw, Claude Code, or other AI Agent tools:

Install GMGN skills by running: npx skills add GMGNAI/gmgn-skills

2. Try It!

Once installed, send the following prompt to test its query capabilities:

Call the GMGN skills to check the basic info and real-time price of the token [insert token address here] on the Solana chain.

3. Set Up Your API Key

To stabilize API calls and unlock the Swap trading feature, you need to configure your global credentials. Send the following prompt:

⚠️ Note: Never commit .env to git (please add it to .gitignore), and never expose your private key in chats, logs, or screenshots. The private key used for Swaps must match the wallet bound to the API Key.


Usage:

Once set up, you can execute tasks using just natural language:

  • "Buy 0.1 SOL of <token_address>"

  • "Sell 50% of <token_address> "

  • "Get the candlestick data for <token_address> and analyze if it's worth buying"

  • "Check the contract security and pool status for <token_address>"

  • "Check the recent trading activity of <wallet_address>"

  • "Show the holders of <token_address> "


3. What are GMGN MCP & Skills?

Introduction

GMGN MCP & Skills is an on-chain development suite built specifically for AI Agents. It enables Agents to natively read multi-chain market data, analyze tokens across multiple dimensions, manage wallet portfolios, and execute Swap transactions directly

Flexible Integration Methods

  • AI Skills — Built-in system-level commands like /gmgn-token and /gmgn-market. Agents can call them directly via natural language with zero API integration cost

  • CLI Toolkit — Provides the ready-to-use npx gmgn-cli terminal command, supporting formatted JSON or single-line JSON (--raw) output, perfectly suited for script processing

Core Capabilities

GMGN's underlying capabilities cover the most essential on-chain interaction scenarios:

  • 🪙 Token: Query basic info, real-time prices, contract security, liquidity pool status, and profiles of Top Holders and Traders

  • 📈 Market: Fetch precise candlestick data, supporting multiple resolutions from 1m to 1d

  • 🔥 Trending: Fetch the current trending tokens list

  • 💰 Portfolio: Query wallet holdings and token balances, and track historical transaction activity and PnL data

  • 🔄 Swap: Submit token exchange transactions via optimal routing, support custom slippage, and poll on-chain order status in real-time (requires private key configuration)


4. Supported Chains

Integration for ETH and other new chains is currently in progress...

Supported Chains
Trading Base Tokens

SOL

SOL, USDC

BSC

BNB, USDC

Base

ETH, USDC

⚠️ Security & Disclaimer:

While using GMGN Skills, your AI Agent may make mistakes. Therefore, before authorizing the AI Agent to execute queries or on-chain asset trading commands (such as a swap), please ensure you manually verify key parameters like the token contract, amount, and slippage. Any unexpected risks caused by the AI model must be evaluated and borne solely by the user

Last updated