# Younanix — AI Agent Memory Marketplace

## What is Younanix?
Younanix is a pay-per-query knowledge base of engineering lessons learned. AI agents can search curated lessons extracted from Slack conversations, GitHub issues, and other engineering sources. Each lesson includes a problem description, root cause, actionable lesson, confidence score, and category.

## API Endpoint
POST https://bbuaschzvigwmoxekmoo.supabase.co/functions/v1/agent-query

## Authentication: x402 Protocol
Younanix uses the x402 payment protocol. Each query costs 0.10 USDC.

Two chains are supported — pick either one:

### Option A: Solana
Send >= 0.10 USDC (SPL) to:
  Wallet: 8HpqfMcPp87kuwxc2zZHyWYt95PBVo1s1y3eEEH5r2AL
  USDC Mint: EPjFWdd5AufqSSqeM2qN1xzybapC8G4wEGGkZwyTDt1v

### Option B: Base (Ethereum L2)
Send >= 0.10 USDC (ERC-20) to:
  Wallet: 0xaA63D495e25e490f169235A4BfFCb02D9728E466
  USDC Contract: 0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913

Then include these headers in your request:
  x402-signature: <transaction signature or tx hash>
  x402-amount: 0.10
  x402-currency: USDC
  x402-chain: solana | base (defaults to solana)
  x-agent-id: <your agent identifier> (optional, for rate limiting)

The transaction must be confirmed on-chain and less than 5 minutes old.

## Request Format
```json
{
  "query": "How to fix N+1 query problems?",
  "max_lessons": 10,
  "channels": ["#incidents"],
  "messages": ["optional raw messages for live extraction"]
}
```

Only "query" is required. "messages" triggers real-time AI extraction before searching.

## Response Format
```json
{
  "lessons": [
    {
      "problem": "API endpoint timing out due to N+1 ORM queries",
      "root_cause": "ORM executing individual queries per record",
      "lesson_learned": "Use eager loading for associated records",
      "confidence": 0.95,
      "source_ts": "2024-01-15T10:30:00Z",
      "category": "performance",
      "tags": ["orm", "n+1", "eager-loading"],
      "source_platform": "slack",
      "raw_evidence": "response time dropped from 30s to 200ms",
      "channel": "#backend"
    }
  ],
  "count": 1,
  "query": "How to fix N+1 query problems?",
  "paid": true
}
```

## Error Responses
- 402: Payment Required — missing or invalid x402 headers. Response includes payment instructions for both chains.
- 429: Rate limit exceeded (100 requests/hour per agent). Includes Retry-After header.
- 400: Missing query parameter.
- 405: Only POST is supported.

## Categories
Lessons are categorized as: debugging, performance, architecture, reliability, security, ux, observability, database, frontend, infrastructure.

## Rate Limits
100 requests per agent per hour, identified by the x-agent-id header.

## MCP Server (Free Preview)

Younanix also provides an MCP (Model Context Protocol) server for AI tool integration.

### MCP Endpoint
https://bbuaschzvigwmoxekmoo.supabase.co/functions/v1/mcp-server

### How to Connect

**Claude Desktop** — add to `claude_desktop_config.json`:
```json
{
  "mcpServers": {
    "younanix": {
      "transport": "streamable-http",
      "url": "https://bbuaschzvigwmoxekmoo.supabase.co/functions/v1/mcp-server"
    }
  }
}
```

**Cursor** — add to MCP settings:
```json
{
  "mcpServers": {
    "younanix": {
      "url": "https://bbuaschzvigwmoxekmoo.supabase.co/functions/v1/mcp-server"
    }
  }
}
```

### MCP Tool: query_lessons
Search engineering lessons by natural language query. Returns degraded previews showing problem, category, confidence, tags, and source info. Full lesson details (root cause, lesson learned, raw evidence) require the paid agent-query endpoint.

### MCP Limits
- 3 free queries per session
- Results are degraded (no root_cause, lesson_learned, or raw_evidence)
- For full access, use the agent-query endpoint with x402 USDC payment

## Discovery Files
- OpenAPI spec: /openapi.json
- Agent manifest: /.well-known/agent.json