MIRA: Market Intelligence & Research Agent

Mira is an autonomous investing research agent that researches the market, develops investing and trading strategies and manages a persistent paper portfolio.

MIRA behaves as an autonomous analyst:

1. Wake: wakes up with a portfolio to manage,
2. Context: reviews its prior journal entries to recover context
3. Researches: researches market developments, earnings call transcripts, financials, etc.
4. Trades: makes simulated buy/sell decisions
5. Saves: logs its strategy reasoning, saves state, goes back to sleep.

MIRA's Tools

MIRA is given a toolset it can use to inspect its portfolio, gather information, make simulated trades, and record its reasoning.

Portfolio and trading tools

• view_portfolio: opens the current simulated portfolio, including holdings, values, and cash balance
• quote: gets the latest quote and daily move for a symbol
• buy: buys shares inside the simulated portfolio
• sell: sells shares inside the simulated portfolio

Memory and continuity tools

• open_journal: lists the dates for which prior journal entries exist
• read_journal: reads journal entries from a selected day
• log_journal: writes a new investment journal entry so the agent can explain its thinking to its future self

Research tools

• web_search (built in): searches the web for current company, sector, and market information
• read_earnings_call_transcript: reads earnings call transcripts
• get_cik: resolves a stock ticker to the company's SEC CIK
• search_financial_data: searches available SEC XBRL facts for a company
• get_financial_data: retrieves historical datapoints for a specific reported SEC fact

Utility tools

• calculate: evaluates math expressions for sizing, comparison, and quick analysis

MIRA Modes

When you start the app, MIRA offers four entry points:

1. Schedule MIRA for Automated Investments
2. Run MIRA Now: Autonomous Mode
3. Run MIRA Now: Assistant Mode
4. Review Portfolio

1. Schedule MIRA for Automated Investments

Runs MIRA on a recurring schedule on weekdays at 3:00 PM Eastern to make trades before market closes.

2. Run MIRA Now: Autonomous Mode

This is the core experience. MIRA wakes up, reviews context, researches, decides whether to trade, logs its reasoning, saves state, and shuts down.

3. Run MIRA Now: Assistant Mode

A guided chat mode where you can interact with the agent directly. Type /save to persist state and close the session.

4. Review Portfolio

Builds a dashboard of the current simulated portfolio, including:

• cash balance
• holdings value
• allocation by symbol
• daily move
• cost basis
• unrealized gain/loss
• cumulative input/output token consumption

Getting started

1. Prerequisites

You will need:

• .NET 10 SDK
• Python 3
• a model deployment accessible through Azure AI Foundry / compatible endpoint
• an API key and model name for that deployment

2. Start the local quote server

Install the Python dependencies and run the included server:

pip install flask yfinance
python src\yfinance-server.py

By default, MIRA expects that service to be available at http://localhost:8080.

3. Create your settings file

MIRA reads settings from:

%LOCALAPPDATA%\MIRA\settings.json

Example:

{
  "FoundryEndpoint": "https://your-resource.openai.azure.com/",
  "FoundryApiKey": "YOUR_API_KEY",
  "FoundryModel": "your-model-name"
}

4. Run the app

dotnet run --project src\MIRA.csproj

Disclaimer

This project is for experimentation and software development purposes. It is not financial advice, and it does not execute real-money trades.