FinquillOS Platform Guide

FinquillOS is not a chatbot. It is a multi-agent research operating system — a network of nine specialized AI agents, each owning a distinct domain of financial intelligence, coordinated by an orchestration layer that routes every conversation to the right expert (or experts) automatically.

When you type a question, it doesn't go to a single model. It passes through the Router Agent, which classifies your intent, selects the right specialist, and — for complex multi-domain requests — fires multiple agents in parallel before synthesizing their outputs into a single coherent response. Long-running tasks (deep research, portfolio reviews) are dispatched as background jobs so you can keep working while the analysis runs.

The result is a research environment that knows your portfolio context, remembers your behavioral patterns across sessions, and coordinates multi-step analysis that no single model call could produce. Think of it as the Bloomberg Terminal rebuilt for the agentic AI era — not a tool you query, but an infrastructure layer that works alongside you.

Every message in Full Desk mode passes through a two-tier hierarchy. The Router Agent sits at the top and dispatches to the specialist agents below. Each specialist has a dedicated tool set and deep context for its domain.

Routing Patterns

• Direct dispatch — Single-domain query goes straight to the right specialist. “What are NVDA's gross margins?” → Research Analyst.
• Parallel delegation — Multi-domain query fires multiple agents simultaneously. “Prepare me for NVDA earnings” → Calendar Strategist + Research Analyst in parallel, outputs synthesized.
• Plan / approve / execute — Complex multi-step request triggers a plan card. You review and confirm before the Router sequences the steps.
• Background task dispatch — Deep research and portfolio reviews run as persistent Trigger.dev jobs. Progress is visible in real time; you can keep chatting.

AI Studio is the primary interface for the agent network. Every module in the platform feeds into it — your watchlist state, journal history, calendar events, and wallet positions are all available as live context. Open it anywhere with ⌘ /.

Operating Modes

Research Depth Levels

In Smart and Full Desk modes, the depth selector controls how aggressively the agents use web research tools.

Model Selection

AI Studio supports 10+ model providers through a unified gateway. Models marked with a tools icon support function calling — required for Smart and Full Desk modes. Reasoning models (marked with a brain icon) operate in Simple mode only; they produce higher-quality analytical output but cannot call tools. Groq models are unlocked by adding your API key under Settings → API Keys.

Context-Aware Next Steps

After every response, the agent surfaces 2–4 context-specific follow-up actions as clickable buttons — generated by the agent based on what was just discussed. These are not generic menus; they are inferred next moves specific to your current conversation. Clicking one dispatches the suggestion directly back to the agent.

Composed Surfaces

When your ask doesn't fit a standard card — cross-asset comparisons, sector screens, bespoke multi-section briefings — the agent builds a custom visual surface on the fly: KPI grids, tables, charts, callouts, and interactive controls composed into a purpose-built panel. Composed surfaces save to Research automatically, so they're shareable as /r/[slug] URLs and referenced by future agent conversations like any other artifact.

• Refine in place — the panel's “Refine this artifact” input sends follow-ups that update the same surface instead of spawning duplicates. Step through versions with the ‹ › arrows in the panel header.
• Version history is preserved — every time the agent revises a surface, the prior version is captured before the update. Nothing is silently overwritten, and your surfaces build a real revision trail over time.

Slash commands load expert workflow blueprints into the agent on demand. Type /in any AI Studio or chat input to open the skill library — a fuzzy-searchable dropdown of 52 analysis methodologies (DCF, Comps, Kelly Criterion, Buffett's Owner Earnings, FOMC scenarios, and more). Selecting a skill injects its workflow into the agent's instructions for the next message; multiple skills can be active at once and the planner will integrate them into a single coherent run.

How It Works

• Type / to open the dropdown. Keep typing to fuzzy-filter by skill ID, name, tag, or description. ↑↓ to navigate, Enter to enable.
• Enable + send in one shot — /dcf AAPL enables the DCF skill and immediately sends “AAPL” with it active.
• Active skills appear as removable pills near the input. Stack multiple skills — the plan generator weaves them into one integrated workflow (e.g. /dcf + /comps + /bias-detector for a full valuation pass).
• Skills + chains that shaped a response appear as read-only pills above each assistant message — so you can always tell which methodology was applied, not just which ones you loaded. Chain names render in insight color, standalone skills in primary.
• Contextual suggestions appear at the top of the dropdown — skills the platform infers are relevant to what you're currently discussing.
• Available everywhere the agent listens: AI Studio (Smart and Full Desk modes) and the chat panels on Watchlist, Research, Journal, Calendar, Media, and Trading pages.

Skill Library

52 skills organized into 15 categories. The dropdown groups them by category; here's a tour of what's available.

Worked Examples

Command Palette Actions (⌘ K)

Every agent in FinquillOS builds and maintains a structured working memory profile for your account. This is not a chat transcript — it is a persistent JSON schema that captures who you are as an investor and how you work. It persists indefinitely across sessions, devices, and conversations.

What's Tracked

• Risk tolerance — Conservative / moderate / aggressive, and whether you trade options or leverage
• Trading style — Swing, position, day — how you actually use the platform, inferred over time
• Sectors & themes — The markets you monitor and care about, updated as you add tickers and read research
• Behavioral patterns — Detected from your Journal — recurring rule violations, emotional biases, process adherence score
• Research context — Active theses, investment cases in progress, recent research sessions
• Preferred depth — Your default research depth level, inferred from usage patterns

Semantic Recall

Beyond the structured profile, agents perform semantic recall over your research artifacts and journal entries using vector search — surfacing relevant past work even when you don't explicitly reference it. If you wrote a thesis on $MSTR three weeks ago and ask about Bitcoin treasury strategy today, the agent retrieves it.

The Dashboard is your mission control — a cross-agent synthesis of market state, portfolio posture, and upcoming events. It renders on load and refreshes throughout the day.

• Major index performance (S&P 500, NASDAQ, Russell 2000, BTC) at a glance
• Watchlist highlights — top movers, biggest % changes, unusual volume
• AI morning briefing — Generated by the Dashboard Agent: synthesizes overnight moves, analyst upgrades/downgrades, earnings surprises, and macro data releases into a concise, actionable paragraph
• Latest signals from your AI agents, sorted by urgency and convergence score
• Urgent actions bar — surfaces items that need your decision: workflow approval requests, ready-to-act watchlist entries, and priority risk signals. Click any chip to expand details; approvals deep-link straight to the Task Center queue.
• Upcoming calendar events for watchlisted tickers (earnings dates, ex-dividend, splits)
• Quick-access cards for Journal, Research, and AI Studio

The Watchlist is the platform's live context spine — every agent that does portfolio-aware analysis reads from it. Add equities and crypto in the same list; the platform handles data sources and symbology automatically.

• Real-time price data, daily change %, and volume for every ticker
• Analyst price targets, consensus ratings, and recent rating changes
• Sort and filter by performance, sector, asset class, or custom tags
• Click any ticker to open a detailed view with charts, fundamentals, and AI analysis panel
• Attach thesis notes to individual tickers — the Watchlist Manager references these in future analysis
• Structured thesis claims with a one-click Check claims against latest data button. Each claim has an invalidation condition; running the check fires the thesis-monitor task and posts drift warnings to your notifications.
• Every ticker row has an Open full view CTA that deep-links into the Asset Hub — per-asset chat, inline profile editing, linked research artifacts
• Portfolio-level opportunity scoring: which positions have the highest conviction-to-current-price divergence

The Research module is for structured, saveable, shareable investment analysis. Every output is a Research Artifact — a typed document (thesis, analysis, due diligence, trade idea) with structured fields, a conviction score, and a permanent URL.

• Research Analyst workflow: profile generation → SEC filing analysis → analyst consensus pull → deep web research synthesis → thesis formulation → conviction scoring. Watch each step execute in the live tool call timeline.
• Three research depths: Quick (~1 credit), Research (~2), Deep Dive (~5–8). Choose based on decision stakes.
• Every artifact auto-generates Bull Case / Bear Case / Invalidation Criteria / Conviction Score fields
• Save artifacts to collections; attach them to watchlist tickers for persistent context
• Composed views — the custom surfaces agents build in AI Studio land here as artifacts too, refinable in place with every prior version preserved
• Share artifacts publicly — they become /r/[slug] URLs visible on the Explore page
• Fork community research: any public artifact can be forked and edited as your own base

Signals is your centralized feed of proactive AI intelligence — alerts generated by agents without you asking. It surfaces what you should know before you know to ask for it.

• Risk escalations, unusual volume, pattern breaks, and earnings-proximity alerts for watchlisted tickers
• Convergence scoring — signals that are flagged by multiple independent agents simultaneously receive a higher priority score. A volume anomaly spotted by the Media Analyst and confirmed by the Watchlist Manager becomes a high-urgency signal.
• Act on any signal: click “Discuss” to open AI Studio with the signal pre-loaded as context, or “Research” to trigger a deep dive immediately
• Filter by signal type (volume, price, news, technical), priority, or ticker
• Proactive signals run on automated schedules: weekday market scans at 2 PM ET, morning nudges at 2 AM and 2 PM

The Journal is where you record trades and decisions — and where the AI applies behavioral finance analysis to help you trade better. The Journal Coach agent reads your entries over time to detect patterns you can't see yourself.

• Create entries with rich text, tags, and attached tickers; record entry/exit prices, position size, and thesis
• Voice-to-text capture via microphone — dictate entries in real time while markets are moving
• Journal Coach analysis: pattern detection, rule enforcement against your written trading rules, emotion-tagging (fear, greed, conviction, uncertainty)
• Weekly process review: ask the Journal Coach to score your process adherence, identify recurring errors, and generate a one-action improvement item
• Review historical entries alongside performance data — test whether your thesis held up
• Journal patterns feed into your working memory, so agents proactively reference your behavioral profile

The Calendar tracks every event that can move your positions — and helps you build a strategic posture around each one. The Calendar Strategist agent transforms dates into actionable playbooks.

• Earnings dates for every watchlisted ticker, auto-populated from live market data
• Macro events: FOMC meetings, CPI releases, PCE, jobs reports, Fed speeches
• Add custom reminders — position review milestones, stop-loss check dates, thesis invalidation triggers
• Day, week, and month views; upcoming events surfaced on the Dashboard automatically
• Earnings playbook generation: ask the Calendar Strategist to generate a full pre-earnings briefing — historical beat/miss rate, implied move from options, analyst consensus, bull/base/bear scenarios — for any watchlisted ticker

The Media Feed aggregates financial news and runs it through the Media Analyst agent to extract actionable signals — not just headlines, but classified intelligence.

• RSS-powered aggregation across top financial news sources
• Filter by topic, source, or relevance to your watchlist — ticker-aware filtering surfaces only what affects your positions
• AI-summarized headlines: each article compressed to a one-sentence signal-bearing summary so you scan in seconds, not minutes
• Signal detection: the Media Analyst classifies each article by signal type (earnings catalyst, macro event, sector rotation, risk escalation) and confidence level
• Save any article to your research workspace or send it directly to AI Studio as context for follow-up analysis

The Trading module connects your brokerage account to the agent network. Orders are placed through AI conversations — the Trading Agent runs a structured pre-trade checklist before any order is submitted.

• Connect your brokerage in Settings → Integrations (Alpaca supported; more in roadmap)
• Place equities, crypto, and options orders through natural language in AI Studio
• Pre-trade checklist: before every order, the Trading Agent checks (1) thesis validity, (2) position sizing vs. account, (3) risk/reward ratio, (4) existing exposure, (5) upcoming catalysts. A summary card is shown before confirmation.
• View and manage open positions, pending orders, and P&L directly in the Trading module
• Safety confirmations required for: orders over a configurable size threshold, short positions, options contracts, and any trade that conflicts with your journal rules

The Wallet module provides a unified view of your Web3 portfolio — native balances, ERC-20 tokens, and DeFi positions — integrated into the same context layer as your equities watchlist.

• Connect multiple wallets via WalletConnect (300+ supported wallets)
• Track balances and ERC-20 holdings across Ethereum, Arbitrum, Base, and Optimism
• DeFi position tracking: Uniswap V3 liquidity positions (tick ranges, fee accrual, IL exposure) and Aave V3 lending/borrowing (APY, health factor, liquidation thresholds)
• Portfolio view aggregated, by wallet, or by network — switch between them in the header
• Batch transaction builder for atomic multi-step operations (e.g., borrow + swap + supply in one sequence)
• Safety controls: configure max transaction size limits, daily spending caps, and contract allowlists to prevent accidental large transfers

Explore is the community discovery hub — a public surface for the platform's collective research output. The Feed is the social layer for following individual investors.

Explore

• Browse featured research artifacts shared by the community — searchable by ticker, type (thesis, analysis, trade idea), author
• Baskets: curated thematic portfolios built by users and Finquill analysts. Each basket has a factor profile (growth/value/momentum/quality), holdings list, and performance summary. Fork any basket as a starting point for your own.
• Community agent configurations: browse and fork agents built by other users in Agent Lab
• Every public artifact is a shareable URL — /r/[slug] for research, /b/[slug] for baskets

Feed

• Social timeline of public activity from investors you follow: new watchlist adds, published research, and shared trade ideas
• Like and fork shared content — forked research creates a copy in your workspace you can edit and extend
• Filter by content type: watchlist activity, research artifacts, or trade ideas
• Follow investors from their Explore profiles to populate your personalized feed

Agent Lab is the platform's agent builder — create custom AI agents with specific instructions, tool access, and behavioral constraints. Build a specialist for your exact workflow and invoke it from AI Studio.

• Configure agents with a system prompt, name, icon, and description
• Select from 14 named tool groups: finance, research, watchlist, calendar, journal, media, edgar, trading, crypto, alpha-intelligence, and more
• Define custom workflows: chain tool calls in a fixed sequence for recurring analysis tasks
• Save agents and invoke them from AI Studio's Custom Agent mode — your agents appear alongside the built-in specialists
• Share agent configurations publicly or use community presets from the Explore page

Task Center is where every piece of background work the agents do becomes visible, auditable, and re-runnable. Long-running jobs (deep research, portfolio reviews, graph workflows) dispatch here so you can keep working while they run — and come back to inspect the output whenever you're ready.

• Start workflow launcher — one-click invocation for the five graph workflows: Due Diligence, Earnings Playbook, Earnings Autopsy, Thematic Basket, and Morning Trade Plan. Each opens a pre-trade plan card with a human approval gate before it finalizes.
• Pending approvals queue — workflows that have drafted a proposal and need your sign-off. The Urgent Actions bar on the Dashboard deep-links here via its Review & Approve chip.
• Background + proposed + scheduled tabs — three feeds in one place: in-flight background jobs, agent-proposed tasks awaiting your acceptance, and the weekly platform sync schedule (with “Run now” buttons for thesis-monitor, weekly-review, and the other on-demand workers).
• Click any row to view the result — every completed row expands inline with the full markdown output, plus a View in Research link when the task produced a saveable artifact. The “View result ↓” chip flags rows that have output waiting for you.
• Status chips — running / completed / failed / needs-approval / stuck, with automatic reconciliation of zombie workflows whose Trigger.dev runs never wrote back.

Press ⌘ K from any page to open the command palette. Navigation chords use two keystrokes: press G then the destination letter.

Global Shortcuts

Navigation Chords (G + key)

In-App Interactions

Credits meter the raw AI cost of what you run: 1 credit = $0.001 of model and tool spend, multiplied by your tier's markup (Free 2.5×, Plus 1.2×). Navigation and reading are free; every AI turn, briefing, workflow and background task is metered on the tokens it actually used, and the counter is visible in the AI Studio toolbar. The ranges below are measured on the current default model and quoted as Free-tier · Plus credits.

A capped action launches only when your balance covers its whole cap; you are then charged what the run actually spent, never more than the cap. Free accounts get a one-time 500-credit grant that does not renew — enough for one red-team or one earnings autopsy. Plus is $12/mo and includes 5,000 credits each month at the lower markup; credit packs top up any tier and never expire.

Accessible from Dashboard → Settings or via G S.

• Profile — Name, email, avatar, and investor profile (risk tolerance, trading style) that seeds agent working memory
• Integrations — Connect brokerage (Alpaca), add API keys for third-party model providers (Groq for BYOK models)
• Notifications — Configure which signal types trigger alerts and via which channels
• Appearance — Light / dark theme toggle
• Security — Two-factor authentication, active session management
• Billing — Subscription plan, credit balance, purchase credit packs