AlphaMeta

AlphaMeta is an algorithmic trading platform that unifies five surfaces in a single process:

• REST API gateway — 80+ commands for market data, SEC EDGAR fundamentals, options analytics, trading, portfolio, automation, and watchlists, served over HTTP by a FastAPI gateway and callable via curl or any HTTP client.
• Real-time streaming — a WebSocket endpoint (/ws/stream) broadcasting live ticks, quotes, and 1m–1h bars at tick rate, plus custom formula streams.
• Backtesting engine — Rust-based strategy backtests with QuantStats/Plotly reports, available via the alphameta backtest CLI, REST API, Jupyter magics, and the web UI.
• Live strategy hosting — hot-reloadable strategy files managed over REST, running against live market data.
• Embedded MCP server — an MCP tool layer (/mcp) exposing the full command surface to AI agents: market data, fundamentals, portfolio, and live trading. Trading is opt-in and gated behind a three-step confirm-token flow with risk limits and audit logging.
• AI agent harness — a built-in chat server that runs agents locally and plugs into any LLM, orchestrating MCP tools end-to-end from natural language.

Broker backends: IBKR (stocks, options, futures, crypto) and QMT (China A-shares). Also includes SEC EDGAR research commands — filing listing, 13F investors, insider trades, XBRL financial statements and segment data.

Quick start

uv tool install alphameta    # or: pip install alphameta

Get a free API key at alphameta.app, then create ~/.alphameta/.env:

# API Key (Required) — get one free at https://alphameta.app
ALPHAMETA_API_KEY=am-xxx

# IBKR Account ID (Required for trading)
ALPHAMETA_IBKR_ACCOUNT_ID=U1234567

# IBKR Gateway connection
ALPHAMETA_IBKR_HOST=127.0.0.1
ALPHAMETA_IBKR_PORT=4001
ALPHAMETA_IBKR_CLIENT_ID=0

# Timezone (default: US/Eastern)
# ALPHAMETA_TIMEZONE=Asia/Shanghai

Start the server (choose a broker backend: ibkr or qmt):

alphameta --broker ibkr

Open the web UI in your browser (server must be running first):

alphameta web

The API serves at http://127.0.0.1:18080 by default:

# Live quotes
curl -X POST "http://127.0.0.1:18080/api/v1/execute" \
  -H "Content-Type: application/json" \
  -d '{"cmd": "quote AAPL"}'

# SEC research (no IBKR needed)
curl -X POST "http://127.0.0.1:18080/api/v1/execute" \
  -H "Content-Type: application/json" \
  -d '{"cmd": "filings AAPL 10-K"}'

Real-time streaming

Connect to the WebSocket endpoint for push-based market data instead of polling:

ws://127.0.0.1:18080/ws/stream

The streamer broadcasts live ticks and quotes (throttled to tick rate), real-time bars (1m / 5m / 15m / 30m / 1h), and evaluated formula streams, with heartbeat and connection-state events.

Backtesting & strategies

Run backtests from the CLI with QuantStats or Plotly reports:

alphameta backtest --strategy my_strategy.py --symbol SPY --report quantstats --benchmark SPY

The same engine is available over REST (/api/v1/backtest, /history, /{id}/report), in Jupyter via alphameta.research (get_pricing / run_backtest) and IPython magics, and in the web UI backtest workbench with equity charts and run history.

Live strategies are hot-reloadable Python files in ~/.alphameta/strategies, managed over REST (/api/v1/strategies — list / start / stop / reload / logs).

MCP server (AI agents)

An embedded MCP server (streamable HTTP) is mounted at /mcp and enabled by default. It exposes the full command surface — quote, kline, chain, positions, balance, orders, gex/dex, maxpain, calendar, news, SEC filings, and more — to any MCP-compatible agent.

Trading tools are available but opt-in (ALPHAMETA_MCP_TRADING_ENABLED=true, default off); when enabled, every trade goes through a three-step confirm-token flow: trade_preview → trade_confirm → trade_transmit, with risk limits, symbol blacklists, and an audit log.

The built-in chat server doubles as a local agent harness: it can plug into any LLM provider and orchestrates these MCP tools end-to-end from natural language.

AI Agent Skills

alphameta-skills provides agent skills for AI coding assistants (Claude Code, Codex, Cursor, etc.) — use AlphaMeta in natural language for market data, trading, portfolio analysis, and more.

npx skills add intelliscale/alphameta-skills -g

Features

Commands by category

Prerequisites

1. Broker backend — pick one with --broker:

   • ibkr — Interactive Brokers (requires IB Gateway or TWS, see below)
   • qmt — QMT for China A-shares

2. Interactive Brokers Account (for ibkr): You need an active IB account

3. IB Gateway or TWS (for ibkr): Download and install either:

   • IB Gateway (Stable) — Recommended for API-only use
   • IB Gateway (Latest) — Latest features
   • Trader Workstation (TWS) — Full trading platform

4. API Configuration (for ibkr):

   • Enable API access in TWS/Gateway: Configure → API → Settings and check "Enable ActiveX and Socket Clients"
   • Set appropriate port (default: 7497 for TWS, 4001 for Gateway)
   • Add 127.0.0.1 to trusted IPs if connecting locally

Examples

Bulk symbol expansion:

# Add/remove multiple option strikes at once
curl -X POST "http://127.0.0.1:18080/api/v1/execute" \
  -H "Content-Type: application/json" \
  -d '{"cmd": "add SPY240412{P,C}005{1,2,3}0000"}'

# Remove by row id
curl -X POST "http://127.0.0.1:18080/api/v1/execute" \
  -H "Content-Type: application/json" \
  -d '{"cmd": "remove :{31..37}"}'

Multi-symbol orders:

# Buy ~$15k of each symbol
curl -X POST "http://127.0.0.1:18080/api/v1/execute" \
  -H "Content-Type: application/json" \
  -d '{"cmd": "expand buy {META,MSFT,NVDA,AMD,AAPL} $15_000 MID"}'

Flexible order pricing:

# By quantity or dollar amount, with bracket orders
curl -X POST "http://127.0.0.1:18080/api/v1/execute" \
  -H "Content-Type: application/json" \
  -d '{"cmd": "buy AAPL 100 AF @ 233.33 ± 10"}'

Conditional triggers:

# Execute command when price condition is met
curl -X POST "http://127.0.0.1:18080/api/v1/execute" \
  -H "Content-Type: application/json" \
  -d '{"cmd": "if AAPL last > 300: buy AAPL 100 AF"}'

Backtest a strategy:

# QuantStats report against a benchmark
alphameta backtest --strategy my_strategy.py --symbol SPY --report quantstats --benchmark SPY

Account calculator:

# Calculate shares buyable on margin, combine with growth projections
curl -X POST "http://127.0.0.1:18080/api/v1/execute" \
  -H "Content-Type: application/json" \
  -d '{"cmd": "(/ :BP3 AAPL)"}'

License

© 2026 Intelliscale Technologies, Inc.