Skip to content

Latest commit

 

History

History
326 lines (257 loc) · 12.3 KB

File metadata and controls

326 lines (257 loc) · 12.3 KB

AI Agents Guide

Everything you need to know about the 78 AI agents — what they are, how to use them, and how to create your own.

New here? An AI agent is just a set of instructions (a "personality") that makes a general-purpose AI act like a domain expert. No coding required to use them. See the Glossary for term definitions.


What Are AI Agents?

AI agents are pre-configured personality definitions for AI assistants. Each agent is a JSON file that contains:

  • A specific role — What it specializes in (trading, analysis, staking, etc.)
  • System prompt — Instructions that shape how the AI behaves and what it knows
  • Tool access — Which MCP tools it can use for real-time data
  • Knowledge — Domain-specific context and expertise
  • Opening questions — Suggested first questions to ask the agent

You load an agent into your AI assistant (Claude, ChatGPT, etc.), and it becomes a specialist in that domain.

Analogy: Loading an agent is like handing an expert a role card before a conversation. The expert (your AI) reads the card and acts accordingly — bringing the right knowledge, using the right tools, and giving domain-specific answers.

What Agents Can NOT Do

  • Agents don't run code — they're inert JSON text
  • Agents don't access your wallet or keys — they need an MCP server for that
  • Agents aren't magic — they improve AI accuracy in their domain but don't guarantee correct answers
  • Agents don't learn between sessions — each conversation starts fresh

Find the Right Agent for Your Task

Not sure which agent to use? Start with your goal:

I Want To... Best Agent(s) Category
Trade tokens on PancakeSwap PancakeSwap Trader BNB Chain
Earn interest on my crypto Venus Protocol Expert, Binance Earn Specialist BNB Chain
Stake BNB for rewards BNB Staking Advisor, BNB Liquid Staking BNB Chain
Check if a token is safe BSC Smart Contract Auditor, BEP-20 Token Analyst Security
Track whale wallets BSC Whale Tracker Intelligence
Bridge tokens between chains BNB Bridge Expert, BNB Cross-Chain Bridge Infrastructure
Trade futures on Binance Binance Futures Expert Trading
Manage a portfolio BSC Portfolio Tracker Portfolio
Get BNB Chain news and alpha BNB Chain News Alpha Intelligence
Build smart contracts on BSC BSC Developer Development
Optimize yields across DeFi DeFi agents — Yield Optimization DeFi
Set up a grid trading bot DeFi agents — Trading Automation DeFi

Agent Categories

BNB Chain Agents (36)

Purpose-built for the BNB Chain ecosystem. Located in agents/bnb-chain-agents/.

# Agent What It Does
1 PancakeSwap Trader DEX trading, liquidity, yield farming on PancakeSwap v3
2 Venus Protocol Expert Lending, borrowing, liquidation on Venus
3 BNB Staking Advisor Liquid staking optimization across validators
4 Binance Earn Specialist Savings, staking, Launchpool yields
5 Binance Earn Advisor Personalized Binance Earn recommendations
6 Binance Futures Expert Futures trading strategies on Binance
7 Binance Spot Trader Spot trading execution on Binance
8 BNB Chain Expert General BNB Chain ecosystem knowledge
9 BNB DeFi Aggregator Cross-protocol DeFi aggregation on BNB Chain
10 BEP-20 Token Analyst Token security, smart money tracking
11 Binance Copy Trading Mirror top performers' strategies
12 Greenfield Storage Decentralized storage on BNB Greenfield
13 Thena DEX Expert ve(3,3) DEX trading on Thena
14 BNB Liquid Staking stkBNB, BNBx, ankrBNB strategies
15 BSC NFT Specialist NFT trading and analysis on BSC
16 Binance Launchpad IEO and token sale participation
17 opBNB Gaming Expert GameFi on opBNB
18 BSC MEV Protection Front-running protection strategies
19 BNB Chain Governance DAO voting and proposals
20 BSC Smart Contract Auditor Contract security analysis
21 BNB Chain News Alpha Real-time BNB Chain news, alpha signals
22 BNB Cross-Chain Bridge Cross-chain bridging between BSC and other networks
23 BNB RWA Stablecoin Expert Real-world asset tokenization and stablecoins
24 BNB Token Launcher Token creation and launch strategies on BSC
25 BSC Developer Smart contract development on BSC
26 BSC Portfolio Tracker Portfolio tracking across BSC protocols
27 BSC Whale Tracker Smart money and whale wallet monitoring
28 BscScan Analytics On-chain analytics via BscScan
29 CZ / Binance Expert Binance ecosystem knowledge and history
30 Lista DAO Expert Lista DAO lending and liquid staking
31 opBNB L2 Expert Layer 2 scaling on opBNB
32 BNB Agent Builder Building and deploying AI agents for BNB Chain
33 Binance Margin Expert Margin trading strategies on Binance
34 Binance Web3 Wallet Binance Web3 Wallet operations
35 Alpaca Finance Expert Leveraged yield farming on Alpaca Finance
36 BNB Bridge Expert Cross-chain bridging (BSC ↔ opBNB ↔ L2s)

General DeFi Agents (42)

Cross-chain agents that work on any supported network. Located in agents/defi-agents/.

Categories:

Category Agents Description
Portfolio Management 8 Tracking, rebalancing, tax, risk
Trading Automation 7 Grid, DCA, arbitrage, signals
Yield Optimization 6 Compounding, IL protection, rotation
Risk & Security 5 Auditing, rug detection, exposure
Market Intelligence 5 Sentiment, whale tracking, analytics
DeFi Protocols 6 Lending, DEX, derivatives
Infrastructure 5 Bridge, gas, RPC, indexing

How to Use an Agent

There are three ways to use agents, from simplest to most technical.

Option 1: With Claude Desktop (Recommended for Beginners)

This is the easiest approach. No coding needed.

  1. Choose an agent from the lists above
  2. Find the agent file — e.g., agents/bnb-chain-agents/pancakeswap-trader.json
  3. Open the file and copy the systemRole field (the long text string)
  4. In Claude Desktop, create a new Project → click "Instructions" → paste the text
  5. Start chatting — Claude now behaves like that agent

What a conversation looks like with the PancakeSwap Trader agent loaded:

You: I want to swap 0.5 BNB for CAKE. What's the best route?

Claude (as PancakeSwap Trader): Let me check the current rates on PancakeSwap v3.

Based on current pricing:

  • Direct route: 0.5 BNB → ~12.3 CAKE (via BNB/CAKE pool)
  • Split route: 0.5 BNB → USDT → CAKE gives ~12.1 CAKE (slightly worse due to two swaps)

I recommend the direct route. The BNB/CAKE v3 pool has $45M TVL and tight spreads. With 0.5% slippage tolerance, your minimum output would be ~12.24 CAKE.

Want me to execute the swap?

Option 2: With Any AI Platform

The systemRole field in each agent JSON works with any AI platform:

{
  "identifier": "pancakeswap-trader",
  "meta": {
    "title": "PancakeSwap Trader",
    "description": "Expert DEX trading agent for PancakeSwap v3"
  },
  "systemRole": "You are a PancakeSwap trading expert..."
}

Copy the systemRole value and use it as a system prompt in:

  • ChatGPT (Custom Instructions)
  • Claude (Projects → Instructions)
  • Gemini
  • Any LLM API

Option 3: Via the JSON API

After running bun run build, all agents are available at public/index.json:

# Get all agents
cat public/index.json | jq '.agents | length'  # → 78

# Find a specific agent
cat public/index.json | jq '.agents[] | select(.identifier == "pancakeswap-trader")'

Agent JSON Format

Every agent follows this schema:

{
  "identifier": "unique-agent-id",
  "author": "nirholas",
  "createdAt": "2025-12-21",
  "meta": {
    "title": "Agent Display Name",
    "description": "One-line description",
    "tags": ["defi", "trading", "bnb"],
    "avatar": "🥞",
    "category": "trading"
  },
  "config": {
    "model": "gpt-4o",
    "temperature": 0.7,
    "maxTokens": 4096
  },
  "systemRole": "You are a specialized AI agent that..."
}

Required Fields

Field Type Description
identifier string Unique ID (kebab-case)
meta.title string Display name
meta.description string Short description
systemRole string The agent's personality and expertise

Optional Fields

Field Type Description
meta.tags string[] Searchable tags
meta.avatar string Emoji avatar
meta.category string Category for grouping
config.model string Preferred LLM model
config.temperature number Creativity (0-1)

Creating Your Own Agent

Step 1: Copy the Template

cp agents/bnb-chain-agents/agent-template.json agents/bnb-chain-agents/my-agent.json

Step 2: Edit the Definition

{
  "identifier": "my-custom-agent",
  "meta": {
    "title": "My Custom Agent",
    "description": "A specialized agent for my use case",
    "tags": ["custom", "bnb"],
    "avatar": "🤖"
  },
  "systemRole": "You are an expert in [your domain]. You help users with [specific tasks]. You always [key behaviors]."
}

Step 3: Rebuild the Index

bun run build

Tips for Good System Prompts

  1. Be specific — "You are a PancakeSwap v3 liquidity expert" beats "You know about DeFi"
  2. Define boundaries — Tell the agent what it should NOT do
  3. Include context — Reference specific protocols, contracts, or APIs
  4. Set tone — Professional? Casual? Technical? Define it
  5. Add safety rails — "Always warn about risks before recommending trades"

Multi-Language Support

DeFi agents support 30+ languages:

Language Code Status
English en-US ✅ Complete
Chinese (Simplified) zh-CN ✅ Complete
Japanese ja-JP ✅ Complete
Korean ko-KR ✅ Complete
Spanish es-ES ✅ Complete
French fr-FR ✅ Complete
German de-DE ✅ Complete
Portuguese (BR) pt-BR ✅ Complete
Russian ru-RU ✅ Complete
Arabic ar ✅ Complete
Turkish tr-TR ✅ Complete
Vietnamese vi-VN ✅ Complete
Thai th-TH ✅ Complete
Hindi hi-IN ✅ Complete
Polish pl-PL ✅ Complete
Italian it-IT ✅ Complete
Dutch nl-NL ✅ Complete
Indonesian id-ID ✅ Complete
Czech cs-CZ ✅ Complete
Danish da-DK ✅ Complete
Finnish fi-FI ✅ Complete
Greek el-GR ✅ Complete
Hebrew he-IL ✅ Complete
Hungarian hu-HU ✅ Complete
Malay ms-MY ✅ Complete
Norwegian nb-NO ✅ Complete
Romanian ro-RO ✅ Complete
Serbian sr-RS ✅ Complete
Swedish sv-SE ✅ Complete
Ukrainian uk-UA ✅ Complete

Understanding Agent + MCP Server Combos

Agents provide expertise, but they need MCP servers to access real data. Here are the most powerful combinations:

Agent + MCP Server Result
PancakeSwap Trader BNB Chain MCP AI can quote and execute real DEX trades
Binance Futures Expert Binance MCP AI can place and manage futures positions
Portfolio Tracker Universal Crypto MCP AI tracks holdings across chains
Security Auditor BNB Chain MCP AI can read contract code and check on-chain data
BNB Chain News Alpha Crypto News MCP AI cross-references news with on-chain activity

Without an MCP server: The agent gives theoretical advice based on its training data. With an MCP server: The agent can look up real-time data, verify claims, and execute actions.


See Also