TradeAgent 1.0.0

TradeAgent 1.0.0, the first stable release.

Install: pipx install tradeagent, then follow the Quick start.

• Dashboard: tradeagent dashboard opens a local web view with today's schedule, every AI decision and its result, per-arm research progress, and live trades. English and 简体中文, light and dark.
• Autopilot: a daily Paper run in which Codex, quant signals and a random pick are scored side by side.
• Live trading: one command per trade on Robinhood's official agentic API, with hard caps and an automatic exit.
• Options (calls and puts): coming soon.

Full changes: CHANGELOG.md.

TradeAgent 1.0.0-beta.9

First release on PyPI: pipx install tradeagent

Highlights

• tradeagent login — sign in to the official Robinhood Trading MCP in your browser; TradeAgent never sees your password.
• tradeagent doctor — offline checklist of what is set up and what to run next.
• Robinhood API updates are now checked for compatibility: additive changes keep working, breaking changes stop trading with a clear reason, and recover can still exit.
• Grouped help, short human summaries, per-user default config.
• README rewritten, with a Simplified Chinese version.

Beta: no strategy has a proven edge yet. Not financial advice. Full notes in CHANGELOG.md.

TradeAgent v1.0.0-beta.1

TradeAgent's first packaged beta release combines prospective market research with a
separate engine for explicitly owner-authorized execution.

Key features

• Market scanning, immutable forecasts and cost-aware quantitative TradePlans.
• Scheduled SHADOW comparisons of Quant Only, Codex Only, Quant + Codex and seeded
  Random, with separate evidence pools and coverage reporting.
• Manual Codex-assisted research using the repository's opportunity analyst skill.
• Owner-authorized US equity/ETF execution through the official Robinhood Trading
  MCP, with deterministic risk checks, durable order identity, reconciliation and recovery.
• Synthetic research demos and a simulated broker entry/exit lifecycle.

Scheduled SHADOW workers never review, place or cancel orders. LIVE requires a
separate explicit owner request, private local configuration, external authentication
and current broker authorization.

Installation

Download tradeagent-1.0.0b1-py3-none-any.whl and SHA256SUMS from this release.
Verify the wheel's SHA-256 against SHA256SUMS, then:

python3 -m venv .venv
source .venv/bin/activate
python -m pip install ./tradeagent-1.0.0b1-py3-none-any.whl
tradeagent demo --demo-dir work/demo
tradeagent inspect --state-dir work/demo
tradeagent run-once --paper --state-dir work/paper

The wheel contains the CLI, broker contracts and owner configuration example.
The source archive contains scripts, tests and documentation. For the Codex skill,
clone the tagged repository:

git clone --branch v1.0.0-beta.1 https://github.com/danielye0010/TradeAgent.git

Python package version: 1.0.0b1. Git tag: v1.0.0-beta.1.
This release is distributed through GitHub and is not published to PyPI.

Supported platforms

Linux with Python 3.12–3.14, including Linux under WSL2. Scheduled research requires
systemd user services. Native Windows and macOS are not supported. Keep operational
state on local Linux ext2/ext3/ext4, btrfs or xfs storage; Windows mounts, network
shares and tmpfs are rejected.

Current limitations

• Prospective profitability and Codex uplift are not established. Historical
  results, synthetic tests and economic eligibility screens do not establish a
  repeatable trading edge.
• SHADOW outcomes use modeled costs and prices; they are not broker fills. Missing
  data and unavailable assessments remain explicit.
• Market access requires an external owner-managed OAuth helper. Scheduled Codex
  assessment requires an authenticated compatible Codex CLI and an available host.
• Owner LIVE execution supports US long equities/ETFs. Options and short selling
  are outside that path. Experimental LIVE is separately requested and does not
  require established edge.
• Broker permissions, data freshness, host availability and market liquidity can
  block entries or exits. Automated exit logic cannot guarantee a fill.

See the README,
SHADOW setup
and LIVE guide.
Trading involves risk of loss.