An AI hedge fund research manager that continuously investigates investment opportunities, forms evidence-based theses, and communicates insights autonomously via email.
AEON is a persistent AI research agent that operates in a continuous loop: gather context, plan research, execute tool calls, analyze findings, communicate insights, sleep strategically, repeat. It does not execute trades — it researches and recommends.
The system is built around cost-aware autonomy. Every LLM token and API call has a measurable cost, tracked against a daily budget.
- 31 research tools across market data, web intelligence, analysis, communication, and memory
- LLM-first architecture — all reasoning, planning, and recommendations flow through the LLM
- Multi-provider LLM support — Ollama (local, free) and AWS Bedrock (cloud)
- Persistent consciousness — SQLite-backed memory with findings, theses, recommendations, and steering inputs
- Real-time consciousness stream — live feed of agent thinking, categorized as
[PLANNING],[RESEARCH],[FINDING],[RECOMMENDATION],[TOOL_CALL],[SLEEPING],[STEERING] - Interactive TUI — terminal interface with live streaming and steering input
- Email reports — professional HTML research updates sent to your inbox
- User steering — direct the agent's focus via CLI, TUI, or email replies
git clone https://github.com/akshaylakkur/AEON.git
cd AEON
bash install.shThe installer walks through LLM provider selection, email configuration, search providers, market data sources, and budget setup.
# Launch interactive TUI (default)
aeonctl
# Start as background daemon
aeonctl start -d
# Send steering input
aeonctl steer "Focus on AI semiconductor stocks"
# Watch the agent think in real time
aeonctl log -f
# Check research status
aeonctl status
# View findings and recommendations
aeonctl history
# Show configuration
aeonctl config
# Stop the agent
aeonctl stopcp .env.example .env # Edit with your credentials
docker compose up -d
docker compose logs -fAEON/
├── aeon/ # Main package
│ ├── __init__.py # Version
│ ├── __main__.py # python -m aeon entry point
│ ├── app.py # AEON class (thin wrapper)
│ ├── cli.py # CLI (aeonctl)
│ ├── core/ # Event bus, state machine, config, consciousness, neural orchestrator
│ ├── cortex/ # LLM providers, reasoning engine, tool registry
│ ├── tools/ # 31 research tools (market, web, analysis, communication, memory)
│ ├── senses/ # Data connectors (market data, intelligence, sentiment)
│ ├── limbs/ # Output interfaces (email, IMAP, notifications)
│ ├── ledger/ # Cost tracking, P&L, burn analysis
│ ├── analytics/ # Alpha generation, risk management, backtesting
│ ├── security/ # Vault, spend caps, sandboxing, audit trail
│ ├── metamind/ # Self-analysis, adaptation, strategy journal
│ ├── reflexes/ # Circuit breakers, API health monitoring
│ ├── simulation/ # Simulation framework for testing
│ ├── tui/ # Interactive terminal UI (Textual)
│ └── orchestrator/ # Main orchestrator, user communication layer
├── tests/ # Comprehensive test suite
├── aeonctl # CLI launcher (bash fallback)
├── install.sh # Installation wizard
├── pyproject.toml # Package config
├── requirements.txt # Dependencies
├── Dockerfile # Production container
├── docker-compose.yml # Docker Compose
├── .env.example # Configuration template
└── data/ # Runtime data (created on first run, gitignored)
User Steering (optional, via CLI/TUI/email)
│
▼
PLANNING — LLM decomposes objective into 2-6 subtasks
│
▼
RESEARCHING — Execute one tool call per turn, LLM decides next action
│
▼
ANALYZING — Evaluate findings, generate investment theses
│
▼
COMMUNICATING — Compile and email professional research report
│
▼
SLEEPING — Cost-aware rest (shorter during market hours with findings)
│
└──→ repeat
All configuration is done via environment variables (.env file). See .env.example for the full list.
Key settings:
AEON_LLM_PROVIDER—ollama(local, free) orbedrock(AWS cloud)AEON_GUIDANCE_PROMPT— What to research (your investment thesis)AEON_RESEARCH_BUDGET_DAILY— Daily LLM inference budget in USDAEON_SEARCH_PROVIDER—duckduckgo(free),serpapi, orbrave- SMTP/IMAP settings for email communication
- Python 3.11+
- OS: macOS or Linux
- LLM: Ollama (local) or AWS Bedrock credentials
.venv/bin/python -m pytestTests use pytest-asyncio with asyncio_mode = "auto". External services are mocked.