PlugWiki

gmgn-cooking Skill: Launch Meme Coins on Pump.fun and Other Launchpads

Lets your agent launch a token on a launchpad such as Pump.fun or FourMeme through the gmgn-cli, and read launchpad creation stats.

Installs
16.7K
on skills.sh
Stars
636
on GitHub
Updated
Oct 2026
last repo commit

Ask it to create a token or cook a coin and the agent walks you through a guided flow: chain and launchpad, name, symbol, logo, initial buy amount and optional fee and auto-sell settings. Then it runs gmgn-cli cooking create. A second command, cooking stats, reports how many tokens each launchpad created.

This spends real money and cannot be undone. Creating a token needs a GMGN API key and a private key for local signing, both kept in a file on your machine. Read the amount and chain back before you confirm.

Good for

  • Meme coin creators who want to launch from the terminal
  • Developers building agent workflows on Solana, BSC or Base
  • Researchers who only need launchpad creation stats

Install gmgn-cooking

Claude Code

  1. 1

    Copy this prompt.

    Install the gmgn-cooking skill in Claude Code. Run this command:
    
    npm install -g gmgn-cli
    
    Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

The README installs the whole GMGN skill set. Put the keys in ~/.config/gmgn/.env.

From the vendor's docs. Source.

What's in the skill

When it loads

[FINANCIAL EXECUTION] Create and launch meme coins and crypto tokens on launchpads (Pump.fun, FourMeme, Bonk, BAGS, Flap, Klik, Clanker, etc.) via bonding curve fair launch, or query token creation stats by launchpad via GMGN API. Use when user asks to create a token, launch a meme coin, cook a coin, deploy on a launchpad, or check launchpad creation stats on Solana, BSC, or Base.

Sections in SKILL.md

  1. Core Concepts
  2. Financial Risk Notice
  3. Sub-commands
  4. Supported Chains
  5. Supported Launchpads by Chain
  6. Quote Token conversion (when --raised-token is set)
  7. Prerequisites
  8. Rate Limit Handling
  9. Credential Model
  10. cooking stats Usage
  11. cooking stats Response Fields
  12. cooking create Parameters
  13. Advanced API Fields
  14. Platform capability matrix
  15. Pump.fun (--dex pump)
  16. Bonk (--dex bonk)

Key facts

Author
GMGN
Installs on skills.sh
16,715 (Nov 2, 2026)

What it includes

  • gmgn-cli config --check before any command

    If it fails, the agent runs gmgn-cli config and applies the API key you send it.

    Runs code · Config check

  • Token creation counts per launchpad

    Needs only the API key.

    Runs code · cooking stats

  • Deploys a token and makes the initial buy

    Needs API key and private key. Takes chain, launchpad, wallet address, name, symbol, buy amount, image and slippage. Anti-MEV works on Solana only.

    Runs code · cooking create

  • Eight steps from chain choice to execution

    Collects platform, name, symbol, logo, buy amount, optional details, platform modes and auto-sell, then runs.

    Guided launch flow

  • Sell configs, bonding-curve and fee-share settings

    Platform-specific fields for Pump.fun, Bonk, BAGS, Flap and FourMeme.

    Write · Auto-sell and fee options

  • gmgn-cli order get until confirmed

    Polls every 2 seconds for up to 30 seconds. The new token address is in report.output_token.

    Runs code · Order polling

From the official docs

Permissions

High risk

What it can do

Watch out for

  • It spends real money and cannot be undone

    cooking create deploys a token contract and makes the initial buy. Check chain, launchpad, wallet and amount before the agent runs it. github.com

  • No separate final confirmation step

    The guided flow ends with "execute the request directly" once the fields are collected. Auto-sell configs sell automatically after launch, and fee-share splits permanently route token revenue to the listed accounts, per SKILL.md. github.com

  • Your private key sits in a plain file

    The setup writes GMGN_PRIVATE_KEY to ~/.config/gmgn/.env. Use a dedicated wallet with small funds, and do not paste the key into chat. github.com

  • A project .env can override your keys

    The CLI reads a .env in the working folder after the global one. Make sure a stray .env does not point at a different wallet. github.com

  • The README quick start makes the agent generate and store keys

    It has the agent run openssl to write an Ed25519 private key to /tmp/gmgn_private.pem, asks you to send back the API key in chat, then writes keys to ~/.config/gmgn/.env with chmod 600. It also says swap capability needs your egress IP whitelisted. github.com

  • It may run network diagnostics

    On 401 or 403 it runs ifconfig or ip addr and requests ipv6.icanhazip.com to test IPv6. github.com

Technical details
Turning on read-onlyUse only cooking stats, which needs the API key but not the private key.
Talks toGMGN OpenAPI over HTTPS. https://ipv6.icanhazip.com (only when diagnosing IPv6)
Sign-in and keysKeys are stored in ~/.config/gmgn/.env; a .env in the current folder overrides it.
Scopes
  • GMGN_API_KEY

  • GMGN_PRIVATE_KEY (token creation)

Questions

What does gmgn-cooking do?

It lets your agent create a token on a launchpad through the gmgn-cli, and check creation stats. Creation deploys a real token contract and makes an initial buy.

How do I install it?

Use the install box above. You need the gmgn-cli, the skill and your own API key from gmgn.ai/ai. Creating a token also needs a private key, so set that up only if you mean to launch.

Does it need an API key?

Yes. Stats need GMGN_API_KEY. Creating a token also needs GMGN_PRIVATE_KEY, and the key must belong to the wallet bound to the API key. The README says a public demo key exists for read-only testing only.

Can I undo a launch?

No. The skill says every create deploys an on-chain contract, spends the initial buy, and cannot be reversed. Use only funds you are willing to lose.

Does it ask before it spends money?

Only partly. SKILL.md says to confirm the human-readable buy amount with you and to ask for each field one at a time, but its last step (Step 8) says to execute directly once the fields are collected. There is no separate final yes/no step in the file. If you want one, tell the agent to show the full command and wait before running it.

Who pays, and is there a test mode?

You pay. The initial buy and network fees come from the wallet bound to your GMGN API key (SOL, BNB or ETH depending on chain). Optional bundle and sniper buys spend from the extra wallets you list. SKILL.md describes no testnet or dry-run for cooking create. The README's public demo key is for read-only testing only, so use cooking stats to check your setup and treat every create as real.

Is it safe?

It can spend real funds from your wallet, and it keeps your API key and private key in a plain env file on your machine. The skill says the private key is used for local signing only and is never passed as a command argument. Read every parameter before you approve.

Why do I get 401 or 403 with a correct key?

The skill points to IPv6 first. gmgn-cli only works over IPv4, so it tells you to disable IPv6 on your network interface.

Does it work in Codex, Cursor or Cline?

The README documents Codex CLI (clone the repo and symlink its skills folder into ~/.agents/skills, shown in the install box), OpenCode (same method), Cursor (skills are found through the repo's .cursor-plugin folder, then restart Cursor) and Cline (point its Skills directory at the installed package's skills folder). The README gives no Claude app upload route; the skill also needs the gmgn-cli installed on your machine.

Information

updated Nov 2, 2026

Folder
skills/gmgn-cooking

gmgnai/gmgn-skillsskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2
  1. GMGN skills README · github.com
  2. gmgn-cooking SKILL.md · github.com

Updated Nov 2, 2026.