tradingcodex  by monarchjuno

Agentic investment workflow harness for Codex

Created 1 month ago
273 stars

Top 94.2% on SourcePulse

GitHubView on GitHub
Project Summary

This project provides an agentic harness for Codex-native investment workflows, enabling research, thesis review, portfolio management, and execution checks. It targets users who want to leverage AI for structured investment processes, offering a benefit of auditable, safe, and service-gated integration rather than autonomous trading.

How It Works

TradingCodex acts as a durable operating system for AI agents, coordinating work between Codex and specialist agents with bounded judgments. It enforces a structured workflow from evidence gathering through analysis, valuation, risk review, and finally, service-gated execution. The architecture separates AI coordination (Codex) from durable functions like policy, order management, and auditing (Django), utilizing Django templates, HTMX, and plain JavaScript for a lean frontend without build steps. This approach prioritizes safety through explicit service gates and independent judgment review.

Quick Start & Requirements

  • Primary install / run command: From an empty workspace, run: uvx --refresh --from tradingcodex tcx attach . && ./tcx doctor .. Restart Codex and open the generated workspace to load configurations.
  • Non-default prerequisites: Requires a Codex-native environment. Python version must be >= 3.11 and < 3.15.
  • Links: Detailed information is available in the docs/ directory. installation.md covers persistent CLI install, GitHub-main install, and service details.

Highlighted Details

  • Agentic workflow planning transforms natural-language prompts into staged plans with defined roles and states.
  • Independent judgment review by a dedicated agent challenges accepted artifacts for contrary evidence and overconfidence.
  • Decision Packages encapsulate investment theses with forecastable claims, assumptions, and contrary evidence.
  • Strategy skills allow users to define reusable rules without granting execution or broker authority.
  • Research memory is file-native, storing artifacts as ordinary workspace files.
  • Execution is service-gated, requiring explicit approvals and policy checks before submission.

Maintenance & Community

The project has shipped core features and is currently at version 0.3.1, focusing on enhanced validation and richer provider profiles. Future plans include verified adapters and managed services. Contributions are accepted under Apache-2.0 with DCO sign-off. No specific community links (e.g., Discord, Slack) are provided in the README.

Licensing & Compatibility

The project is licensed under Apache-2.0, covering source code, templates, and documentation. The TradingCodex name, logos, and official marks are not included in this license. Apache-2.0 generally permits commercial use and linking.

Limitations & Caveats

TradingCodex is not an autonomous trading bot; natural language inputs do not directly trigger broker actions. Live broker support requires installed, reviewed providers and explicit gates, with paper execution being the default. The system actively blocks or constrains direct live broker requests, raw API calls, self-issued approvals, and other high-risk operations. It explicitly states it does not provide financial, investment, legal, tax, or regulatory advice.

Health Check
Last Commit

2 days ago

Responsiveness

Inactive

Pull Requests (30d)
1
Issues (30d)
2
Star History
274 stars in the last 30 days

Explore Similar Projects

Feedback? Help us improve.