Documentation

VIGIL × WinnyWoo is one AI agent across two surfaces: the VIGIL workspace (council, meetings, studio, vault, finance, CRM, mail) and the WinnyWoo trading desk. This guide explains what each does and how to begin.

Create your account

Sign up with email, or continue with Google, Apple, GitHub, or Railway. Email signups confirm via a link. Your session is shared across both surfaces — sign in once.

First moves

• Convene a council in the Meeting Room over a decision you're weighing.
• Open the Trade Desk to see live candles and the signal debate (read-only until you connect a broker).
• Add a few transactions in Finance, or a contact in the CRM, to see the agent reason over real data.
• Drop a document in the Vault so drafts and answers cite your actual material.

AI Council & Meeting Room

Convene a panel of advisor lenses (CFO, CTO, Legal, Product) over a topic or meeting transcript. They argue, score each other, and a chairman synthesizes a verdict with a readiness score. Dissent is deliberate — you see the bear case, not just the bull.

Trade Desk

Live OHLCV candles, model forecasts, and a data-grounded signal debate (bull/bear/risk) enriched from multiple market sources. The agent proposes trades; it never sizes or places them on its own.

Finance

A books/ledger: capture a transaction, classify it into a category, reconcile it. A running P&L and by-category rollup the CFO methodology reasons over.

CRM

Contacts and a deal pipeline (lead → qualified → proposal → negotiation → won/lost). Weighted pipeline value rolls up to Finance; deals can route to the Council for review.

Mail

Inbox triage over your mailbox (via the himalaya transport when connected). The agent classifies messages into buckets and priorities. Outbound is review-then-send — drafts are prepared, never auto-dispatched.

Studio

Draft, refine, and crystallize artifacts — proposals, briefs, contracts, memos, reports. It brainstorms the shape first, then drafts only the approach you approve, grounded in your Vault.

Vault

Your real documents. The agent reads and cites them rather than guessing what a contract, invoice, or statement says.

Audit

A tamper-evident, hash-chained log of what the agent and you did — every decision, proposal, and approval, defensible after the fact.

WinnyWoo is connect-your-own-broker. You add your exchange API keys from your account dashboard — they're stored against your account and used only to act on your behalf.

• Without a broker connected, the desk is fully usable in read-only / markets mode — forecasts, signals, and the debate.
• With a broker connected, the agent can propose orders sized against your real NAV — still behind the approval gate.
• Keys are never shared between users and never used to act for anyone but you.

Every order the agent proposes must pass a single-use human approval before it can execute. This is a hard rule the agent cannot bypass.

• Each approval is single-use — it authorizes exactly one order, then is consumed.
• Position size is capped at 5% of your NAV by default.
• The agent proposes and explains; you approve, modify, or reject. Nothing moves money without you.

Your data is scoped to you. Every record — transactions, deals, mail, artifacts, documents, broker credentials — is isolated per user and enforced at the database layer.

• Row-level security plus an additional gateway-side scope guard on every read and write.
• The agent only ever sees the data of the signed-in user it's working for.
• You can export or delete your data from your account settings.

Is this financial advice?

No. VIGIL × WinnyWoo provides information and tooling, not licensed financial, tax, or legal advice. You make the decisions.

Will it trade automatically?

No. The agent proposes trades; a human approval gate stands between any proposal and execution, capped at 5% of NAV.

Where do my broker keys live?

Against your own account, used only to act for you. They are never shared across users.

What if no AI key is configured?

LLM-backed features (Studio drafting, Mail triage, the Council) degrade to a safe deterministic stub rather than failing — the rest of the workspace still works.

Do the advisors always agree?

Deliberately not. The council surfaces dissent and a readiness score, and you can always override the verdict.