AgentHalo is a collection of LLM-powered AI agents running inside TEE (Intel TDX) for Private, Verifiable AI.
AgentHalo can interpret natural language trading strategies, analyze real-time market signals, and autonomously execute trades across multiple chains.
All operations can be cryptographically verified using TEE remote attestation off-chain, and on-chain as well via zero knowledge proof of the attestation for every operation happening inside the TEE.
For more information on on-chain verification of TEE remote attestation, please refer to TEE Remote Attestation On-chain Verification
- LLM-powered agents capable of processing complex, unstructured signals for trading decisions
- Intelligent tool selection and chaining for complex multi-step analysis
- Dynamic composition and execution of Python code using available tools
- Natural language strategy definition and real-time reasoning
- Iterative agentic reasoning to evaluate market conditions, weigh multiple input signals, and make trading decisions given input trading strategy
- Real-time strategy execution and monitoring
- Execution modes:
- Automated trading alerts via Telegram
- Autonomous trade execution
- Multi-chain support with DEX integrations (Uniswap, Jupiter)
- Python 3.11 or higher
- Download and install Python
- Poetry 2.1 or higher
git clone https://github.com/denniswon/agenthalo.git
cd agenthalo# For basic installation
make install
# For development (includes dev dependencies)
make dev-installNote: Poetry manages its own virtual environments, so a separate virtual environment should not be required.
-
LLM API Key:
- Anthropic API Key if using Claude models (default)
- OpenAI API Key if using GPT models
- or any other LLM provider supported by LiteLLM
-
Blockchain Access:
- Alchemy API Key for blockchain data, or another RPC provider of choice
-
Optional - Telegram Bot for notifications:
- Create a bot through BotFather with
/newbotand securely save the chat id and the bot token
- Create a bot through BotFather with
cp .env.example .envLLM Configuration (at least one required):
ANTHROPIC_API_KEY: Your Anthropic API key if using Claude models (default)OPENAI_API_KEY: Your OpenAI API key if using GPT models
Blockchain Access:
ALCHEMY_API_KEY: Your Alchemy API key for accessing blockchain data
Notification settings:
TELEGRAM_BOT_TOKEN: Required for sending alerts via Telegram botTELEGRAM_CHAT_ID: Required chat ID for receiving Telegram alertsTELEGRAM_SERVER_IP: IP address for Telegram server (defaults to 0.0.0.0)TELEGRAM_SERVER_PORT: Port for Telegram server (defaults to 8000)
Logging configuration:
LOG_LEVEL: Sets logging verbosity level (defaults to INFO)LOG_FORMAT: Custom format for log messages (default: "%(asctime)s - %(name)s - %(levelname)s - %(message)s")
We also need to download the DStack simulator:
# Mac
wget https://github.com/Leechael/tappd-simulator/releases/download/v0.1.4/tappd-simulator-0.1.4-aarch64-apple-darwin.tgz
tar -xvf tappd-simulator-0.1.4-aarch64-apple-darwin.tgz
cd tappd-simulator-0.1.4-aarch64-apple-darwin
./tappd-simulator -l unix:/tmp/tappd.sock
# Linux
wget https://github.com/Leechael/tappd-simulator/releases/download/v0.1.4/tappd-simulator-0.1.4-x86_64-linux-musl.tgz
tar -xvf tappd-simulator-0.1.4-x86_64-linux-musl.tgz
cd tappd-simulator-0.1.4-x86_64-linux-musl
./tappd-simulator -l unix:/tmp/tappd.sockOnce the simulator is running, you need to open another terminal to start your FastAPI development server:
# Start the FastAPI dev server
python -m fastapi dev# Run all tests
make all-tests
# Run specific test suites
make unit-tests
make integration-tests # Requires API keys to be specifiedThe project uses:
- Black for code formatting
- isort for import sorting
- ruff for linting
- mypy for type checking
# Format code
poetry run black .
poetry run isort .
# Run linters
poetry run ruff check .
poetry run mypy .or use Makefile shortcuts:
make dev-lint