ZenInvest
An open agentic investment system for governed decisions.
ZenInvest is an open-source investing system that brings screening, research, debate, and execution into one governed workflow. A three-model committee — Claude on strategy, GPT-4o as skeptic, Gemini on risk — researches and argues each idea with live tools, a proactive macro layer sets the market regime, and deterministic Python controls keep final veto power over capital. It runs autonomously cycle-to-cycle, but every cycle is pausable, auditable, and gated by hard rules no model can override.
featured
Product framing
Retail investors face information overload, fragmented workflows, and too many opaque tools. ZenInvest is designed to reduce that burden without pretending judgment can be outsourced.
Capability 6,900+ US equities screened
Capability 3-model investment committee
Capability 1,341 automated tests + fail-closed promotion gates
Stack
Why ZenInvest
A technical product, not just another market feed.
ZenInvest combines adversarial agent design, tool use, deterministic controls, and operator oversight in one transparent system.
Dynamic tool use lets agents pull live evidence during reasoning
Proactive macro intelligence sets the market regime (RISK_ON / RISK_OFF / NEUTRAL) for each cycle
Deterministic veto logic keeps hard safety rules above model output
Operator interfaces and logs keep each cycle inspectable
How it works
A structured loop from screening to review.
Each cycle is staged so outcomes remain inspectable, governed, and testable.
Screen
Scans a broad tradable universe of 6,900+ US equities and builds a focused opportunity set.
Debate
A multi-agent committee reviews each candidate from research, skeptical, and risk-first viewpoints, rebutting each other before a verdict.
Decide
Deterministic risk logic applies hard constraints and can veto any decision.
Execute and review
Execution, alerts, journals, and evaluation close the loop, while a refresh lane keeps the system synced to broker truth between cycles.
Differentiation
Built as an inspectable agentic system.
The emphasis is on structured challenge, clear risk handling, and inspectable behavior rather than black-box promises.
Agentic architecture that stays inspectable
Built in public so the logic and trade-offs can be inspected rather than hidden behind a proprietary black box.
Operator-supervised autonomy
The system runs autonomously, but every cycle is pausable, auditable, and gated by deterministic checks the models cannot override. Chat-initiated trades require explicit human confirmation.
Learns only behind hard gates
A shadow learning loop turns every decision and outcome into training data, but it stays read-only — it never influences live trades until it clears alpha-adjusted, regime-stratified promotion gates.
Technical product posture over return promises
A technical product experiment with educational value, not a machine for guaranteed returns.
Architecture and ecosystem
A committee-style system backed by a wider tooling ecosystem.
ZenInvest combines orchestration, market data, evaluation, and execution tooling across a broad set of APIs and platforms.
Core architecture
Multi-agent orchestration, dynamic tool use, deterministic guardrails.
The product combines market intelligence, explicit skepticism, and deterministic rules that no model can override. The goal is not to replace judgment, but to give it a better operating system.
Ecosystem
Use cases
Designed for serious operators, not passive spectators.
Useful for investors and builders who want structure, transparency, and control.
Retail investor co-pilot
For people who want structured research support without surrendering judgment to a black box.
AI-native trading laboratory
For builders exploring adversarial agent design, governed execution, and transparent evaluation.
Operator command center
For users who prefer a dashboard-led workflow that unifies research, risk, decisions, and outcomes.
Governance and operations
Safety rules and operator visibility are first-class, not bolted on.
The system enforces deterministic controls and maintains transparency across runs, decisions, and reporting.
Safety guardrails
Hard constraints remain above model output.
- Concentration limits for single names and sectors
- Drawdown state machine: ACTIVE, CAUTIOUS, and HALTED
- Cash floor and exposure constraints enforced by policy
- Risk veto cannot be overridden by LLM outputs
- Cost-aware degradation: the pipeline sheds models gracefully before it ever halts
Operational signals
- 1,341 automated tests with fail-closed evaluation gates
- Docker Compose deployment on VPS
- Run history, alerting, and audit trails
- Daily and weekly reporting with cost and outcome visibility
Disclaimer
Public project, not financial advice.
The product can be ambitious while remaining careful about claims and live use.
ZenInvest is not financial advice. It is an educational, research, and product experimentation project currently running on a paper-trading account. Human oversight remains essential, and any live use should be approached cautiously.