worldquant-orchestrator

End-to-end control of WorldQuant BRAIN from any coding agent.
Data discovery → alpha generation → backtesting → analysis → submission,
driven through BRAIN's official REST API rather than the web UI.

Quick start · Install skills on skills.sh · The loop · Safety · Docs

_{Synthetic agent-chat demonstration · 25-second preview. All expressions, identifiers and results shown are illustrative;
this is not a live account recording or evidence of investment performance.
Watch the full 58-second skills demo → · Synthetic walkthrough →}

---

✦ WQO skills

Give your coding agent the skills to research on WorldQuant BRAIN. Describe an idea in chat, and the agent can find data, build expressions, run backtests, compare candidates and explain the submission checks.

The core wqo-cli skill handles setup and the full research workflow. Eight companion skills add guidance for specific tasks. The skills install the WQO CLI automatically when it is missing; you do not need to install WQO separately or clone this repository.

---

⚡ Quick start

From the project where you want to use WQO, install the core skill:

npx skills add aaditya-v-more/worldquant-orchestrator --skill wqo-cli

Choose your coding agent in the installer. Keep installation project-local by omitting --global. To install the core and all eight companions together:

npx skills add aaditya-v-more/worldquant-orchestrator --skill '*'

Use macOS or Linux with Node.js/npm and either uv or Python 3.12+ available. The skill handles WQO installation in your project automatically.

Open your agent chat in that project and ask:

Use the wqo-cli skill to set up WQO and help me connect my BRAIN account.

You need your own BRAIN account. The skill guides you through private credential setup and pauses for you to complete Persona verification if required. Keep your password out of chat.

Then describe your research:

Build an alpha from analyst EPS estimates. Backtest it on USA TOP3000, compare up to five variants, and tell me whether any clear the submission checks. Do not submit anything.

The agent uses the available skills to carry out the request and report the results, including failed candidates and unavailable checks. Live backtests use your BRAIN simulation quota. Submission requires your explicit approval for the specific alpha.

Browse the core skill on skills.sh or see skill setup and troubleshooting.

---

🔄 The loop

flowchart LR
    A(["🔑 auth"]) --> B(["🗂️ data"])
    B --> C(["⛏️ mine"])
    C --> D(["📊 analyze"])
    D --> E{"🚦 gate"}
    E -->|blocked| C
    E -->|clean| F(["🧑 your call"])
    F --> G(["🚀 submit"])

    classDef step fill:#0D1425,stroke:#5B8DEF,stroke-width:2px,color:#E6EDF3
    classDef check fill:#0D1425,stroke:#A78BFA,stroke-width:2px,color:#E6EDF3
    classDef human fill:#0D1425,stroke:#34D399,stroke-width:2px,color:#E6EDF3
    class A,B,C,D,G step
    class E check
    class F human

Loading

Stage	Skill	What it does
🧭	wqo-cli	Automatic setup and the full research workflow
🔑	wqo-auth	Session, quotas, level, biometric handoff
🗂️	wqo-data	Datasets, datafields, operators — cached 7 days
⛏️	wqo-mine	Generate candidates, backtest in batch, rank survivors
🔬	wqo-simulate	One expression, or a sweep over settings and fields
📊	wqo-analyze	IS stats, PnL, yearly breakdown, correlations, checks
🏷️	wqo-alphas	Filter and organize the library — names, tags, colours
🚀	wqo-submit	Full check report, then waits for your explicit yes
🏆	wqo-account	Rank, competitions, teams, events, standing

---

🔒 Safety model

• You approve submission for each alpha. A research request or passing check report is not permission to submit. The submission skill pauses for your explicit yes.
• You handle credentials and identity verification. The agent guides setup and waits for you to complete any Persona check.
• Research stays within your budget. The skills respect account limits and stop when quota is exhausted.
• Results include the failures. The agent reports rejected candidates and unavailable checks alongside promising results. Backtests do not guarantee future performance.

Submission consumes quota and permanently affects future self-correlation. See safety controls and limits for the underlying safeguards and how to handle uncertain outcomes.

---

📚 Reference

Installed skills include their own setup and workflow instructions. For work on this repository, see AGENTS.md and Contributing.

Document	Contents
Portable skills	Installation, automatic CLI setup and troubleshooting
CLI reference	Commands, simulation settings, grades and rate limiting for direct CLI use
Manual CLI installation	Optional standalone CLI setup, upgrades and ledger migration
docs/fastexpr.md	FASTEXPR syntax constraints, the operator set actually available at this account level, and the pv1 field list
docs/alpha-research.md	Local gate rules, fitness arithmetic, sourced examples and research limitations
docs/api-map.md	Every website tab mapped to its endpoint, plus what is not available

The research reference distinguishes local defaults from platform checks, links the original publications and labels unverified assumptions. No source PDF is bundled with this repository.

Caution

There is no leaderboard endpoint. GET /alphas is 405 and /leaderboards is 404: you can read your own rank, never other people's alphas. This is deliberate — alphas are contributor IP.

📁 Repository layout

docs/
  alpha-research.md   distilled research reference
  api-map.md          tab -> endpoint map, and what is unavailable
  fastexpr.md         expression syntax and operator availability
wqo/
  endpoints.py    API URLs
  config.py       paths, pacing, budgets, gate thresholds
  session.py      authenticated session, cookie persistence, Persona handoff
  pacing.py       interval throttle, Retry-After, backoff
  store.py        SQLite ledger (audit trail + budget accounting)
  catalog.py      datasets / datafields / operators, cached
  simulate.py     submit, poll, normalize; adaptive slot manager
  alphas.py       records, PnL, yearly stats, correlations, properties
  gate.py         local thresholds + BRAIN checks -> PASS/FAIL report
  submit.py       confirmation-gated submission
  mining/         templates, candidate generation, ranking
  account.py      competitions, standing, team, learn, notifications
  __main__.py     CLI

---

Contributing · Code of conduct · Security · Licence

_{MIT licensed. Not affiliated with or endorsed by WorldQuant.
Alpha expressions are sent to BRAIN for simulation; local ledgers and account statistics should stay out of Git — see CONTRIBUTING.md.}

↑ Back to top