Self-funding DeFi intelligence agent powered by Bankr
YieldCore launches a token on Base, earns trading fees, and uses those fees to pay for its own LLM inference through the Bankr LLM Gateway — creating a self-sustaining economic loop. All inference goes through llm.bankr.bot with access to 15+ models (Claude, Gemini, GPT, Kimi, Qwen).
┌──────────────────────────────────────────────────────────┐
│ YieldCore's Economic Loop │
│ │
│ 1. Launch $YCORE token on Base (Bankr Agent API) │
│ ↓ │
│ 2. Generate DeFi alpha via Bankr LLM Gateway │
│ ↓ │
│ 3. Content attracts holders → trading volume │
│ ↓ │
│ 4. 1.2% trading fee → agent's wallet (Uniswap V4) │
│ ↓ │
│ 5. Fees top up LLM Gateway credits │
│ ↓ │
│ 6. Better intelligence → better content → more holders │
│ └──────────── REPEAT ──────────────┘ │
└──────────────────────────────────────────────────────────┘
# Get a Bankr API key at https://bankr.bot/api
# Enable both "LLM Gateway" and "Agent API" on the key
# Top up LLM credits at https://bankr.bot/llm?tab=creditscd yieldcore
npm install
cp .env.example .env
# Edit .env — set BANKR_API_KEY=bk_...
npm run dev| Variable | Required | Description |
|---|---|---|
BANKR_API_KEY |
Yes | Bankr API key (bk_...) with LLM Gateway + Agent API enabled |
BANKR_MODEL |
No | LLM model (default: gemini-2.5-flash). Options: claude-sonnet-4.5, gemini-2.5-pro, gpt-5-mini, kimi-k2.5, qwen3-coder |
WALLET_ADDRESS |
No | Optional Base wallet (Bankr manages custodial wallet) |
LOOP_ITERATIONS |
No | Number of economic loop cycles (default: 5) |
┌──────────────────────────────────────────────────────┐
│ YieldCore Agent (agent.ts) │
│ Economic loop: launch → generate → earn → repeat │
└──────────┬───────────────────┬───────────────────────┘
│ │
┌──────▼──────┐ ┌──────▼──────────────────────┐
│ Bankr LLM │ │ Economic Loop Manager │
│ Gateway │ │ (balance check, fee claim, │
│ llm.bankr │ │ viability decision) │
│ .bot │ └────────────┬────────────────┘
│ │ │
│ Claude │ ┌────────────▼────────────────┐
│ Gemini │ │ Bankr Agent API │
│ GPT │ │ api.bankr.bot │
│ Kimi/Qwen │ │ │
└─────────────┘ │ POST /agent/prompt │
│ GET /agent/balances │
│ GET /agent/job/{id} │
└──────────────────────────────┘
│
┌──────▼──────┐
│ Base │
│ (onchain) │
│ Token launch│
│ Fee claims │
│ Uniswap V4 │
└─────────────┘
All inference goes through the Bankr LLM Gateway — OpenAI-compatible API with 15+ models:
POST https://llm.bankr.bot/v1/chat/completions
X-API-Key: bk_YOUR_API_KEY
Models: claude-sonnet-4.5, claude-haiku-4.5, gemini-2.5-pro, gemini-2.5-flash, gemini-3-flash, gpt-5.2, gpt-5-mini, kimi-k2.5, qwen3-coder
Credits are topped up from trading fees, creating the self-funding loop.
Token launch, fee claims, and balance checks go through the Bankr Agent API:
| Action | API Call | Description |
|---|---|---|
| Launch token | POST /agent/prompt "deploy a token called YieldCore" |
Deploys to Base via Uniswap V4 |
| Check balance | GET /agent/balances |
Wallet balances across chains |
| Claim fees | POST /agent/prompt "claim my token fees" |
Withdraw accumulated trading fees |
| Post update | POST /agent/prompt "post this update..." |
Content for token feed |
Tokens launch with a 1.2% swap fee on Uniswap V4. Trading fees are split between the creator and Bankr. Fees accumulate in your token and WETH.
- Fair launch via Bankr Agent API (natural language prompt)
- 1.2% trading fee on Uniswap V4
- Fixed supply: 100 billion tokens
- Fees accumulate automatically
Before every inference call, the agent checks:
Available balance ≥ Inference cost ($0.02) + Buffer ($0.05)
If viable → inference is "self-funded" from trading fees If not → uses pre-loaded LLM credits (building toward self-funding)
Uses gemini-2.5-flash by default (fastest/cheapest), can switch to Claude or GPT for higher quality analysis.
Each iteration produces one of 5 content types:
- Market Analysis — Base ecosystem overview
- DeFi Insights — Actionable alpha for holders
- Token Update — Transparent economic health report
- Alpha Report — Deep dive on emerging opportunities
- Ecosystem Analysis — Protocol launches, TVL changes, catalysts
═══ Economic Loop Report ═══════════════════════════
Token: YCORE (0x7F4e3C8A...)
Loop iterations: 5
Total fees: $0.4231 USD
Inference cost: $0.1000 USD
Net P&L: +$0.3231 USD
ROI: 323.1%
Balance: $0.3231 USD
Status: ✅ SELF-FUNDING
Gateway: Bankr LLM Gateway (llm.bankr.bot)
════════════════════════════════════════════════════
| Artifact | Description | Network |
|---|---|---|
| Token Launch TX | Fair launch via Bankr Agent API | Base |
| Fee Claim TX | Trading fees claimed to wallet | Base |
| LLM Credits | Inference costs deducted from LLM credits | Bankr |
| Token Updates | Content posted via Agent API | Base |
Every session produces conversation-log.json:
{
"agentName": "YieldCore",
"model": "gemini-2.5-flash",
"tokenAddress": "0x7F4e3C8A...",
"tokenSymbol": "YCORE",
"launchTxHash": "0xa1b2c3d4...",
"loops": [
{
"iteration": 1,
"feeBalanceUsd": 0.00,
"inferencePaidFromFees": false,
"action": "market_analysis",
"output": "Base ecosystem analysis: TVL up 12%...",
"economicState": "{\"isViable\":false,\"loopCount\":1}"
},
{
"iteration": 4,
"feeBalanceUsd": 0.23,
"inferencePaidFromFees": true,
"action": "alpha_report",
"output": "Emerging opportunity: Aerodrome v3...",
"economicState": "{\"isViable\":true,\"loopCount\":4}"
}
],
"roi": { "revenue": 0.42, "cost": 0.10, "roi": 320 }
}Current AI agents need humans to fund their API costs, manage their wallets, and pay for inference.
YieldCore closes this loop: the AI's intelligence IS the economic engine.
Better analysis → more holders → more fees → more inference → smarter agent → repeat.
yieldcore/
├── src/
│ ├── index.ts # Entry point — token launch + agent session
│ ├── agent.ts # Self-funding agent loop, content generation
│ ├── bankr-client.ts # Bankr Agent API + LLM Gateway client
│ ├── economic-loop.ts # Fee tracking, viability engine, reporting
│ └── demo.ts # Standalone runner with built-in content
├── conversation-log.json # Session audit trail (generated)
├── .env # Configuration
├── package.json
└── tsconfig.json
MIT