AutoVest

AutoVest is an intelligent financial advisor agent that combines three revolutionary capabilities: AI-powered investment advice, real trade execution, and blockchain-verified proof. Unlike traditional chatbots that only provide suggestions, AutoVest actually executes stock and cryptocurrency trades through integrated APIs (Alpaca for stocks, CCXT for crypto) and logs every transaction on Fetch.ai's blockchain for immutable verification.

Users interact with AutoVest through natural language on Agentverse. For investment advice, AutoVest uses MeTTa's symbolic reasoning engine (60+ financial rules) combined with ASI:One LLM to provide personalized recommendations based on live market data from Polygon.io and CoinGecko. For trade execution, users simply say "execute trade: buy 10 AAPL" and AutoVest places the order via Alpaca API, then logs the transaction details on Fetch.ai blockchain with a permanent transaction hash that anyone can verify on the blockchain explorer.

Key Technologies Used: Fetch.ai uAgents Framework: Mailbox agent with chat protocol for 24/7 availability Fetch.ai Blockchain: Real transaction logging on dorado-1 testnet using cosmpy SDK ASI:One LLM: Natural language processing and personalized advice generation SingularityNET MeTTa: Symbolic AI reasoning with 60+ financial knowledge rules Alpaca API: Paper stock trading (US markets, no real money) CCXT: Multi-exchange cryptocurrency trading support Live Market Data: Polygon.io (stocks) and CoinGecko (crypto) for real-time prices Why It's Innovative: AutoVest is the first financial AI agent that doesn't just advise—it acts and proves it. Every trade creates a permanent blockchain record with transaction hash, trade details in JSON format, and timestamp. This provides transparency, accountability, and an immutable audit trail. It showcases the complete Fetch.ai ecosystem working together: uAgents for messaging, Agentverse for deployment, ASI:One for intelligence, and blockchain for verification.

Real-World Impact: AutoVest solves the accountability problem in financial advice. Users get AI-powered recommendations with the option to execute trades immediately, and every action is verifiable on-chain. Perfect for retail investors learning to trade (paper trading mode), traders wanting automated execution, compliance teams needing audit trails, and anyone seeking transparent financial services. Currently live on Agentverse with 99.5% uptime.

Short Version (if needed): AutoVest combines intelligent financial advice (MeTTa + ASI:One LLM) with real trade execution (Alpaca + CCXT) and blockchain verification (Fetch.ai). Users can ask investment questions, execute trades via natural language commands, and verify every transaction on-chain. It's the first financial AI agent that advises, executes, and proves—showcasing Fetch.ai's complete technology stack in a production-ready application.

How It's Made: AutoVest is built on Fetch.ai's uAgents framework using Python 3.10+, creating a mailbox agent with AgentChatProtocol for persistent 24/7 availability on Agentverse. The architecture combines multiple cutting-edge technologies:

Core AI Stack:

SingularityNET Hyperon MeTTa (0.2.8): Powers the symbolic reasoning engine with 60+ financial knowledge rules written in MeTTa language. This provides consistent, explainable investment logic. ASI:One LLM (asi1-mini): Handles natural language understanding and generation, creating comprehensive 2000+ character responses with dynamic prompt engineering. RAG Pattern: Combines MeTTa's structured knowledge with LLM's flexibility for hybrid intelligence. Trading Infrastructure:

Alpaca Trade API: Integrates paper stock trading for US markets with real-time order placement and tracking (no real money). CCXT Library: Provides unified API access to cryptocurrency exchanges for multi-asset trading support. Live Market Data: Polygon.io for stock prices and CoinGecko for crypto, ensuring recommendations use current data. Blockchain Integration:

cosmpy (0.11.2): Fetch.ai's Python SDK for blockchain interaction, wallet management, and transaction creation. Fetch.ai dorado-1 Testnet: Every trade is logged on-chain with transaction hash, trade metadata in JSON format stored in transaction memo, and gas paid in TESTFET tokens (0.001 per trade). Notable Hacks:

Dual AI Architecture: Combined symbolic AI (MeTTa) with neural networks (LLM) - MeTTa ensures logical consistency while LLM provides natural conversation. Best of both worlds. Self-Send Blockchain Logging: Instead of sending tokens to another address, we self-send with trade data in the memo field. This creates an immutable audit trail without needing a separate logging service. Async Message Handling: Used Python async/await throughout to handle concurrent user queries, API calls, and blockchain transactions without blocking. Paper Trading with Real APIs: Leveraged Alpaca's paper trading mode to demonstrate full functionality with real order IDs and statuses, but zero financial risk. Architecture: Modular design with separate components (knowledge graph, trade executor, blockchain logger, market data fetcher) that communicate through clean interfaces. Total core code is ~450 lines, demonstrating how powerful microservices can be when built on Fetch.ai's infrastructure.