Documentation

What FN2 is

FN2 is a chat box. You type a question about a stock, a sector, the economy, or an earnings call, and the AI answers. What makes the answer worth trusting is everything behind the chat box: live quotes, a large library of earnings-call transcripts, Federal Reserve data, the open web, and a backtesting engine. The model decides which of those it needs for your question, fetches the data, and cites it.

You can work two ways. In chat, you drive the conversation and get answers in real time. With an agent, you hand the same work to a schedule: FN2 runs the research on its own and emails, messages, or texts you the result.

See it in action

Here are real questions and the answers FN2 gave. Click any one to open the live chat. No signup needed.

You do not need special syntax. Plain questions work. But a little specificity pays off, because it tells the model which tools to reach for. Name the ticker, the timeframe, and what you actually want to know.

Quickstart

The fastest way to understand FN2 is to ask it something. Here is the path from sign-up to a scheduled agent in about five minutes.

How FN2 works

Here is what happens after you hit send. FN2 is a language model with a research desk attached. When your question arrives, the model decides which tools it needs, runs them, reads what comes back, and writes an answer grounded in the results. You never call a tool by name. You describe what you want, and the model wires up the steps.

The tools are grouped into skills. A skill is a bundle of related tools the model loads when your question calls for it. Asking about an earnings call loads the transcript-search skill; asking for the macro backdrop loads the FRED skill. This keeps the model focused on the right tools for the job instead of juggling all of them at once.

Two things make the output trustworthy:

• Grounding. Figures come from a tool that fetched them, not from the model's memory. If the data is not available, FN2 says so rather than inventing it.
• Citations. The model marks claims with inline [N] references that link to the underlying source. Citations that do not resolve to a real source are dropped before you ever see them.

Talking to FN2

You ask, FN2 answers, and an answer can carry more than just text: interactive price charts, company cards, earnings transcripts, news, macro tables, and prediction-market odds, depending on what you asked.

The welcome screen

Open the chat and you see live charts for the S&P 500, NASDAQ, and Dow, plus a search bar for jumping straight to any ticker. Click an index to open its panel, with the chart, a read on why it moved, and the key numbers for the day. Your recent agent runs sit just below, one click away.

What an answer can contain

• Text formatted with headers, lists, and emphasis.
• Charts you can zoom and re-range (1D through 5Y).
• Company cards with logo, price, and key metrics.
• Earnings transcripts with the relevant passage highlighted.
• News with source, date, and a one-click "Ask AI".
• Macro tables from FRED with year- and month-over-year changes.
• Citations linking each claim to its source.

History and shortcuts

Every conversation is saved. Search past chats by keyword or browse by date, and pick up any one where you left off. A few keys speed things up:

Citations & sources

When an answer leans on data, FN2 marks the claim with a numbered reference like [1]. Click it to open the source: the exact earnings passage, the filing, the news article, or the web page the model read. This is on by default; there is nothing to turn on.

Citations are a grounding safeguard, not decoration. The model proposes them as it writes, and the server checks each one against a real, stored source before saving. Any citation that does not resolve gets dropped, so a number in your answer always points to something you can open and verify.

Models & file uploads

Choosing a model

A dropdown in the chat input lets you pick which AI model answers. They trade off speed against depth: some are quick for everyday lookups, others reason harder for tougher analysis. Which models you see depends on your plan.

Uploading a document

Attach a PDF or text file and ask about it. Click the attachment button or drag the file into the input. FN2 extracts the full text and reasons over it, which is handy for a 10-K, a research note, or a slide deck.

• Supported: PDF and plain text (.txt).
• Up to 20 MB, one file per message.

Tools the AI can use

The model has a set of tools for fetching real data and running real analysis, and it picks them automatically based on your question. You will never type a tool name; you just ask, and the right tool runs behind the scenes.

Here is the full set, with an example of what triggers each and where it is available.

Quotes & charts

The everyday tool. Ask for a price and FN2 pulls the history and renders an interactive chart you can zoom and re-range. Give it a date window, an exact range, or "the last N days", and name more than one ticker to compare them on a single chart.

You can also jump straight to a ticker with a URL:

Technical analysis MAX plan

Ask for technical analysis and FN2 builds a study chart with the standard indicators overlaid: moving averages (SMA and EMA), RSI, MACD, and a volume panel. This is an on-request tool. Everyday price charts stay clean, and the indicators show up only when you ask for them. Technical analysis is available on the MAX plan.

Market screening

When the question is about the whole market rather than one stock, FN2 queries a price-history workspace to find the names that fit: biggest gainers and losers, volume and dollar-volume leaders, the largest movers in a window. It is read-only and answers breadth questions a single quote lookup can't.

Earnings transcripts

FN2 searches a large library of earnings-call transcripts by meaning, not just keywords, so you can find what management actually said about a topic even when they used different words. You can pull quotes, track how a company's story shifts across quarters, or count how many companies are talking about a theme.

You can also open a specific report directly:

Macro & prediction markets

For the bigger picture, FN2 reads Federal Reserve data (FRED) and prediction markets.

Macro snapshot (FRED)

Ask for the macro backdrop and you get a current read on unemployment, inflation, rates, the yield curve, the VIX, credit spreads, consumer sentiment, and GDP, with historical comparisons. You can also pull any individual series over time.

Prediction markets (Polymarket) MAX plan

FN2 reads live and resolved Polymarket markets, so you can fold real-money odds into your view. It is read-only: FN2 reads the markets, it does not trade them. Prediction-market search is available on the MAX plan.

Web & X search

Web search and browsing

For anything happening now, FN2 searches the web, opens pages, and reads them as clean text, including many paywalled and bot-protected sites. It can also pull tables out of a page as structured data. This is how the model handles current events that are not in the markets or transcript data yet.

X / Twitter search MAX plan

FN2 can search recent public posts on X (Twitter) for market chatter and sentiment. It only runs when you explicitly ask about X, Twitter, tweets, or social sentiment, and it filters for quality (verified accounts or posts with links) so you are not reading noise. A search returns up to about 10 recent posts.

Backtesting MAX plan

Describe a trading strategy and FN2 will test it on historical data. The AI writes the strategy as code, runs it in a sandbox, fixes its own errors, and reports back the return, the number of trades, and the win rate. You steer it in plain language; you do not write any code yourself. Backtesting is available on the MAX plan.

Your watchlists

Your saved watchlists (called collections) double as context for the AI. Ask portfolio-aware questions and FN2 reads your real lists. It can also edit them for you: add or remove tickers, create a new list, rename or delete one, all from chat.

Scheduled agents

An agent is the same AI with the same tools, running on a schedule without you in the chair. Point it at a research task and it executes on its own, then delivers the result. Morning briefings, earnings monitors, thesis trackers, macro watchdogs: anything you would ask in chat, you can ask on a timer.

Starting an agent

Open the Agents panel and start from a template or a blank prompt. You can also just describe it in chat and FN2 will set it up:

Templates

• Morning Briefing: a pre-market summary of overnight moves, news, and your watchlist.
• Earnings Monitor: tracks upcoming and recent earnings for the companies you follow.
• Thesis Tracker: watches for developments that support or undercut a thesis you describe.
• Sector Scanner: ranks names in a sector by the metrics you care about.
• Macro Watchdog: keeps an eye on Fed policy, inflation, the yield curve, and other macro signals.

Configuration

Each agent has a prompt, a model, a maximum run time, and a set of tools it is allowed to use. You can label agents to keep them organized, and shared agents respect roles, so a viewer can't edit or delete the owner's agent.

Output

Every run keeps its full transcript, so you can always reopen an agent and read exactly how it reasoned, and an agent can look back over its own past runs to compare against what it found last time. You choose what gets delivered: the full transcript, a summary, just the final message, or a formatted PDF report.

Schedules

New agents default to a recurring weekly run, Monday at 9am in your timezone, with an email to you turned on. Change any of that in the wizard. The scheduling options are:

• Recurring: hourly, daily, weekdays, weekly, or a custom cron schedule.
• Market-aware: run at the open (9:30am ET) or close (4:00pm ET), with an optional offset like "15 minutes before the close".
• Run Once: a single run at a future time, for example the morning a company reports. It shows up as a scheduled card you can cancel before it fires.
• Timezone-aware: every schedule respects your configured timezone.
• End date: optionally stop a recurring agent on a date you set.

Custom schedules use standard five-field cron syntax:

Agent memory

Recurring agents can remember across runs. A monitor records what it reported last time, then on the next run it compares against fresh data and surfaces only what is genuinely new, instead of re-sending the same headline every morning. Memory is per-agent and kicks in when the agent's job calls for it.

You can see what an agent has stored in a chat's Statistics view, under Agent Memory.

Delivering results

When an agent finishes, it can push the result to you over email, Discord, or text. Email is the default and the most fully featured, so this section walks the email path end to end, then covers Discord and SMS and what to do when a notification does not arrive. This is the technical version, for power users wiring up agents.

How an agent email is built and sent

A delivered email is the end of a fixed pipeline. Nothing in it is improvised by the model except the words:

• The run completes. A scheduled or one-off agent run finishes and produces a structured result, not free text.
• A completion hook fires. The agent calls its delivery tool, for a briefing that is send_briefing_email, which carries the result plus the hook's event and condition.
• mcp-hooks forwards it. The hook service checks the condition, resolves the recipient, and posts to the server. It is also where an unsubscribe or a disabled email type stops the send.
• The server renders the template. send_agent_template_email validates the payload, renders one of the fixed layouts, and attaches the unsubscribe headers.
• The provider delivers it. The rendered email goes out through Mailgun from notifications@fn2.ai.

Anatomy of a hook

A completion hook is what turns a finished run into a notification. You attach one under Advanced → Hooks when you create or edit an agent, and an agent can carry several at once. Each hook has four parts:

• Event — when it fires. on_complete runs after the agent finishes a run.
• Condition — a filter on whether to send. success sends only when the run succeeded; a material-change condition stays quiet on a run that found nothing new.
• Action — the channel: email, discord, or sms.
• Template and recipients — a template_id (briefing, earnings, or alert) and who receives it. Leave recipients empty (to:[]) and the server resolves them to your own account address. The model never chooses who gets the mail.

What the email contains

Three layouts cover the agent email types, each rendered from the structured fields the agent filled in:

• Briefing — a headline, a short intro block in the agent's own words, a row of KPIs, a table, labelled sections, and an optional pull quote.
• Earnings — a headline, the quarter's KPIs, a quote from the call, and sections of detail.
• Alert — a tight headline, a one-line summary, and the supporting details.

Every email is structurally validated before it is sent. A payload with nothing renderable, no headline and no table, sections, or KPIs, is rejected at the server rather than delivered. You never receive a blank email: a misfiring agent sends nothing instead of an empty shell.

The header carries quick links, Help, Edit, and New Agent. While you are signed in they jump straight to the agents view. If your session has expired, the Edit and New Agent links sign you back in through a single-use magic link, so you land where you were going instead of on a login wall. You can turn that off under Settings → Notifications.

Turning email on, off, and unsubscribing

There are two independent layers of control over agent email:

• Notification toggles. Each email type can be switched off under Settings → Notifications. A disabled type is dropped at send time, before the provider is ever called.
• Per-agent unsubscribe. Every agent email footer carries an Unsubscribe link with a signed, per-agent token; clicking it silences that one agent, not all of FN2. Standards-compliant List-Unsubscribe and one-click headers are set too, so your mail client's native Unsubscribe button works.

Every agent email also ends with the standing disclaimer: results are for educational purposes only and should not be considered financial advice.

Discord

Post agent results into a Discord channel. The simplest setup is a channel webhook: paste the webhook URL in Settings → Integrations and FN2 posts a rich embed to that channel. If you link the FN2 Discord app, or add the FN2 bot to your server, you can also get results as a direct message or have them posted to a chosen server channel.

Text message (SMS)

Get a one-line alert texted to your phone. SMS is a Pro feature, with a 30-day trial on new accounts. Verify your number in Settings → Integrations first (with consent), then point an agent at it, or simply ask the chat agent to text you. Texts always go to your own verified number, never an address the model chose. Standard STOP / HELP / START replies work, and there is a daily limit of about 20 texts.

When a notification does not arrive

Agent notifications are deliberately quiet, so a missing one is usually working as intended. Check these in order:

• The hook condition held it back. A success condition skips a failed run, and a material-change condition skips a run that found nothing new. This is the most common reason and it is by design.
• The email type is switched off. Check Settings → Notifications.
• You unsubscribed from that agent. The footer link is per-agent; re-enable it from the agent.
• Test the hook. Trigger the agent with Run Once, then check the run in the agents view, it records whether the notification was sent, skipped, or suppressed.

News

The News page is a curated financial feed you can filter by topic: Markets, Economy, Earnings, Crypto, Commodities, Politics, or Tech. Switch between a grid layout (a lead story plus cards) and a list. Each story shows its headline, source, and timestamp, and an "Ask AI" button hands the article straight to chat for analysis. A live ticker bar across the top tracks the major indices.

Market heatmap

The Market page shows an interactive treemap of the S&P 500. Stocks are grouped by sector and industry, sized by market cap, and colored by the day's performance. Click a sector to zoom in, and hover any tile for the price change and peer context.

Macro dashboard

Ask for a macro snapshot in chat and FN2 lays out a dashboard of the key indicators with current values and recent changes, grouped into labor, inflation, growth, rates, and credit. It also matches the present against historical analog periods, to give the current reading some context. Ask about any single indicator to get its full time series.

Sharing & export

Share a conversation or take it with you.

• Public link: generate a view-only link anyone can open, no account needed.
• Email invite: give specific people access by email, and revoke it any time.
• Export: download a chat as a PDF, Markdown, or plain text, or have it emailed to you.

Languages

FN2 speaks eight languages: English, Chinese, Spanish, French, Italian, Japanese, German, and Korean. The whole interface translates, and the AI answers in your chosen language too.

Switch from the globe in the footer or from Settings → Profile → Locale. If you are signed out, FN2 picks a language from your browser automatically. English is always the fallback.

Accounts & sign-in

Sign-in methods

• Email and password with a verified email address.
• Google for one-click sign-in.

Forgot your password

If you signed up with email and password, use "Forgot password" on the login page. You'll get an emailed link to set a new one. Resetting a password signs out your other sessions. Accounts created with Google are pointed back to "Sign in with Google" instead, since they have no password to reset.

Sessions

A sign-in lasts about 8 hours by default, and you can change that in Settings (up to "stay signed in"). If a session expires while you are mid-chat, FN2 sends you to the login page and brings you right back to the same conversation after you sign in.

Settings

Tune FN2 from the Settings page in the user menu.

• Profile: your name, avatar, timezone, and language.
• Appearance: light or dark mode, and chart density (Standard, Detailed, or Maximum), which controls how many data points AI-generated charts use. The default is Maximum.
• Notifications: how and when you get alerts.
• Integrations: connect Discord and verify a phone for SMS. These are the destinations your agents deliver to.
• Privacy: control your data settings.

A Changelog in the account menu lists recent product updates. It stays out of your way unless you opt into the on-sign-in notice.

Portfolio

Keep the tickers you care about in collections, a set of simple ticker lists. Create as many as you like for different strategies, and add or remove symbols by typing them. The editor lives in Settings → Profile → Portfolio.

Collections do double duty as context for the AI. Your agents and chats can read them ("how is my watchlist doing?"), and the AI can edit them for you on request. See Your watchlists for what the AI can do with them.

API access

Manage API keys from the Developer section in Settings. Create keys with a name and a read or write scope, set an optional expiry, and watch usage per key (request counts, tokens, and last-used time). Activate, deactivate, or delete a key at any time.

Authenticate with a bearer token:

The chat endpoint streams responses as Server-Sent Events. Each event carries a content type that tells you whether it is text, a chart, earnings data, news, or another payload:

Plans & billing

The table below loads from the same plan catalog used at checkout, so it always reflects current pricing.

Every plan includes the full data set: earnings transcripts, FRED macro data, real-time market data, and inline source citations. Higher tiers raise your usage limits and add scheduled-agent capacity. The MAX plan also includes the premium tools: technical analysis, backtesting, prediction markets, and X search.

Frequently asked questions

Can FN2 place trades?

No. FN2 is a research tool. It analyzes markets and reports back, but it does not connect to a brokerage or execute orders. Backtests are simulations on historical data, not live trading. Nothing FN2 produces is financial advice.

How accurate is the analysis?

Answers are built from real data with the sources cited inline, so you can check the underlying transcript, filing, or page yourself. Treat the output as research assistance and verify anything you act on.

What data does FN2 use?

Real-time and historical market data, a large library of earnings-call transcripts, Federal Reserve (FRED) macro data, and the open web, where it reads news, SEC filings, and other primary sources on demand. Some tools are reserved for the MAX plan: technical analysis, backtesting, prediction markets, and X (Twitter) search.

Does FN2 remember me between conversations?

A chat remembers everything within that conversation, but a new chat starts fresh. Scheduled agents are different: they keep memory across runs, so a daily monitor can tell what is new versus what it already reported.

Can I upload my own documents?

Yes. Attach a PDF or text file (up to 20 MB) in chat and ask about it. FN2 reads the full document to answer.

What languages are supported?

Eight: English, Chinese, Spanish, French, Italian, Japanese, German, and Korean. The interface and the AI's answers both follow your choice.

Is my data secure?

Data is encrypted in transit and at rest. Your history, watchlists, and agent configurations are private. Payments are handled by Stripe; FN2 never stores your card details.

Can I cancel anytime?

Yes. Cancel from Account settings whenever you like and keep access through the end of the current billing period.

Troubleshooting

Answers are slow

Questions that span many transcripts or pages take longer; FN2 shows what it is working on as it goes. Narrowing the question (one ticker, one timeframe) usually speeds it up.

A chart isn't loading

Charts need JavaScript. If one comes up blank, refresh the page and check that scripts aren't being blocked.

An agent isn't finishing

Agents have a maximum run time. If one keeps timing out, simplify the prompt, narrow its scope, or raise the time limit in the agent's settings.

A file upload failed

Make sure the file is a PDF or TXT under 20 MB, then try again.

Contact

Reach us at:

• Support: support@fn2.ai
• Enterprise: enterprise@fn2.ai
• Feedback: feedback@fn2.ai. We read every message.

We aim to respond within 24 hours, with priority support for Enterprise customers.