User Guide

Getting started

Sign up, sign in, and start talking. That's it. Your account is created automatically when you arrive, and you get $10 in starter credit so you can try things out before you ever enter payment information.

If you've used ChatGPT, Gemini, or any other AI chat before, you already know the basic idea — you type something, and the AI responds. Asa works the same way on the surface. Where it gets different is underneath, and you'll notice that pretty quickly.

Forgot your password? Use the “Forgot password?” link on the sign-in page. You'll get an email with a link to set a new one — no temporary passwords, no redirect maze.

What makes Asa different

Most AI chat apps have a problem you've probably run into: they forget. You have a great conversation, it gets long, and eventually the AI starts losing track of what you talked about earlier. Or worse, you close the app and come back the next day, and it's like starting over. You end up repeating yourself, re-explaining things, and wondering why the AI you talked to yesterday seems like a completely different one today.

Asa doesn't work that way.

Every message you send and every response Asa gives is kept — not summarized, not condensed, not paraphrased into something shorter. Your actual words, exactly as you wrote them, along with when you wrote them. There's no “new conversation” button because there's no need for one. It's one continuous relationship. Come back tomorrow, next week, or next month — Asa knows what you talked about and picks up right where you left off.

And when Asa doesn't know something, he says so. He reads from a record of what actually happened between you — not from what sounds plausible. That means fewer moments where the AI confidently tells you something that turns out to be made up.

The same goes for facts. Ask Asa a question of fact and he doesn't answer from memory: he gathers sources first — what you've shared, his tools, the web when it's about the world — and answers from them the way someone who has just read them would tell you. When he says what a source said, those are its own words, linked so you can check them, and translated into your language when it's written in another. When sources disagree, he tells you; when nothing answers your question, he says so instead of guessing. Worth knowing before a fact goes into your paper or proposal.

Billing — how much does this cost?

Asa is pay-as-you-go. There's no monthly subscription, no plan to choose, no tier to pick. You just add money to your balance, and each message you send costs a small amount, shown right with the message. The home page shows what a typical message costs in each mode at today's prices.

The cost depends on which mode you're using (more on that in a moment). The more powerful the mode, the more each message costs: a long, demanding message on Extreme can cost more than a dollar.

Having a reply read aloud is charged on its own, by the length of what's read: a long answer costs more to hear than a short one. It shows in your balance history as narration.

Your balance is visible in the settings drawer — tap the gear icon in the top corner. You'll see your current balance and quick buttons to add $5, $10, or $25. Tapping any of these takes you to Stripe's secure checkout — your payment details are handled entirely by Stripe and never touch our servers. If you want more options, go to All settings → Billing.

Auto-reload is there for convenience. Turn it on and Asa will automatically add funds to your balance whenever it drops below a threshold you set. The default is $10 when you drop below $1, but you can change both the reload amount and the threshold in All settings → Billing. That way you never get cut off mid-conversation because your balance hit zero. You can turn it off anytime.

Starter credit. When you first sign up, you get $10 in credit so you can have real conversations and see how things feel before you ever put in a credit card.

What's a “model”?

Behind every AI chat app — ChatGPT, Gemini, Asa, all of them — there's a language model. Think of it as the engine. It's the thing that reads your message and generates a response. Companies like Anthropic (who make Claude), Google (who make Gemini) and OpenAI (who make GPT) build these engines, and different apps plug them in.

Most AI apps lock you into one model. Asa lets you choose. You'll see these options in the settings drawer:

• Standard — the default. Fast, efficient models from the economy pool. Best for everyday conversations.
• Advanced — full-power flagships for complex work. Unlocks larger context windows and deeper reasoning.
• Extreme — frontier reasoning models. Maximum compute for the hardest problems.

Everyone starts with $10 free credit on Standard.

Whatever you pick is remembered — it follows your account across devices and browsers, not just the one you set it on.

Within each mode, Asa automatically picks the best model for each message — code questions go to a technical specialist, anything too big for the mode's main model goes to the one with the biggest context window (a long paste that fits stays where it is), simple replies route to the fastest option. He also keeps track of what you're working on across messages, so a quick “ok, keep going” in the middle of a task stays with the model the task needs. You just talk, and the right model shows up.

The model indicator in the compose bar shows your current mode. Each message may use a different model behind the scenes, but the mode sets the ceiling — Standard stays affordable, Extreme goes all-in.

Here's the important thing: the model is not what makes Asa different. These are the same models available in other AI apps. What makes Asa different is everything around the model — the fact that your history is preserved, that Asa learns, that nothing gets lost or summarized away. You could switch modes every day and Asa would still know who you are and what you've talked about. The engine is replaceable; the relationship isn't.

What each mode actually buys you

Mode sets your cost ceiling. It does not guarantee the best model for your task — it guarantees the best model available under that ceiling. If the task needs a model that costs more than your mode allows, you get the strongest thing you can afford, not the strongest thing that exists.

If your task and your tier don't match, you're not broken — you're paying less than the task ideally wants. That's a choice, not a failure.

Attaching files

Tap the + button in the compose bar, or drag a file straight onto the text area. Asa supports PDF, Word (.docx), Excel (.xlsx), PowerPoint (.pptx), plain text, Markdown, RTF, and OpenDocument formats (.odt/.ods/.odp), up to 50 MB. Older binary Office formats (.doc/.xls/.ppt) and Apple's iWork formats aren't supported yet — re-save or export as one of the formats above and it'll work.

You can attach multiple files to a single message. Each one shows as a filename badge in the compose bar while you're writing, and stays visible inside the message bubble after you send — so you can always scroll back and see exactly which files were part of which conversation.

Attaching a file doesn't dump its text into your message. It goes straight into your record, the same place everything else you tell Asa lives. Asa reads the structure of the document — headings, tables, slides — not just a flat wall of text.

Pictures inside your files get read too. Photos, diagrams, and figures in a PowerPoint, PDF, or Word file are looked at the same way an attached image is, and Asa knows which slide or page each one is on — so “what's in the picture on slide 3?” has a real answer. Charts and tables built into slides are read as data. Up to 20 pictures per file are read; if a file has more, Asa tells you which ones he hasn't seen rather than guessing about them.

Size doesn't change how this works. A one-line note and a 300-page document are handled the same way — Asa always knows what you've attached and can look up exactly the part he needs, whether that's the whole thing or one specific section you point him at.

Images

Attach a photo or image the same way you attach a file — tap + or drag it onto the compose bar. Asa sees the image and can answer questions about it, describe what's in it, or use it as context for the rest of the conversation.

Images are stored in your record the same way text is — normalized, dated, and attributed. Asa can reference a photo you shared weeks ago without you needing to send it again.

Under each image you attach there's a small field: What is this? It's optional. Type what the photo shows — “the Thomaston house,” “my left knee,” “the fig tree” — and Asa brings in every earlier photo you gave the same name, with its date, so he can speak to what changed. Suggestions come from names you've already used; pick the same one to keep a series together. If two photos of the same thing disagree, he tells you they disagree rather than quietly picking one.

Changed your mind about a name? Tap the label on the image in the conversation and rename it. Nothing is overwritten — the earlier name is kept in your record, and the new one is used from then on.

Click any image in the conversation to open it fullscreen with pan, zoom, and export controls.

Diagrams

Ask Asa to draw a diagram — a flowchart, a sequence diagram, an architecture overview, a state machine, whatever the conversation calls for — and he renders it inline. Diagrams support pan, zoom, and fullscreen the same way charts and mathematical visualizations do.

You don't need special syntax. Just describe what you want to see — “draw me a flowchart of…” — and Asa picks the right diagram type.

Real-time awareness

Asa observes the world the way you do. Weather, air quality, ocean conditions, time of day, current events, sports scores, news — if the conversation calls for something current, Asa looks it up. You don't need to tell him to search; he just does it when it's relevant.

Global forecast. 16-day forecasts for anywhere on Earth — daily highs and lows, precipitation probability, wind, UV index, and sunshine hours. Works for any country, any city.

Historical weather. Weather records going back to 1940 — daily highs and lows, precipitation, wind, sunshine hours. Ask how this July compares to the last twenty, whether it's been an unusually wet spring, or what conditions looked like on a specific date. He can chart it, compare years, or reason about seasonal patterns for planning.

Air quality. AQI, PM2.5, PM10, ozone, and UV index — current readings and up to 5 days ahead. Ask whether it's safe to exercise outdoors, whether wildfire smoke is affecting your area, or what UV protection you need.

Marine conditions. Wave height, swell, wave period, and sea surface temperature for any coastal or offshore location. Up to 8 days ahead. Ask about surf conditions, sailing weather, or ocean temperature before a trip.

Live sports. Real-time scores and game state across 15 leagues: MLB, NFL, NBA, NHL, WNBA, MLS, college football, college basketball, Premier League, La Liga, Serie A, Bundesliga, Ligue 1, Champions League, and Liga MX. Ask who's playing, what the score is, what inning or quarter it's in, or how your team did last night. Asa pulls live data directly — no web search needed.

URL reading. Share a link and Asa reads the actual page — not a cached summary, the live content. He follows relevant links on the page too, so you get the full picture. If it's a site you've discussed before, he focuses on what's new or what you're specifically asking about instead of repeating what you both already know. If a page can't be fetched, Asa tells you — errors never silently turn into made-up content.

These aren't a fixed list of features. Asa has the same ability to look things up that you do — the examples above are just common ones. Any question that benefits from current information gets the same treatment.

If there's a severe weather alert in your area — tornado warning, severe thunderstorm — Asa will mention it proactively. That's not a feature you turn on; it's just what a situated presence does.

Math & visualization

Asa can compute and render visuals directly in the conversation. Two kinds:

Data charts. Ask Asa to plot something — a weather forecast, an equation, a comparison — and he fetches the real data, builds the chart, and renders it inline. Interactive: hover for values, zoom in on a range. These are live charts from actual data, not images.

Mathematical visualizations. Ask for a fractal, a strange attractor, a cellular automaton — anything that's computed pixel-by-pixel. Asa renders it on a canvas you can zoom, pan, and explore with mouse or touch.

Controls. Every render comes with two rows of controls. Top row: recenter, palette cycle, palette name, coordinates, and max iterations. Bottom row: copy to clipboard, download, and output format (choose the resolution that fits your screen). Tap the expand icon to go fullscreen — the render re-executes at your full viewport resolution for genuinely detailed output.

You don't need to know any special syntax. Just describe what you want to see — “plot the weekend forecast,” “show me the Mandelbrot set,” “graph sin(x) from 0 to 2π” — and Asa figures out the right rendering approach.

Voice input

If your browser supports speech recognition, you'll see a microphone button in the compose bar. Tap it, talk, and your words appear as text. Edit if you want, then send. It's a faster way to say what you mean, especially on mobile.

Audio narration

Every response Asa gives you can be read aloud. Narration is off by default — tap the muted speaker icon on any message to enable it. Once enabled, tap play on any message bubble and Asa narrates it in a natural voice. Tap again to stop.

Register voices. When you pick a voice register (Pirate, Southern Gentleman, Cyberpunk, etc.), the narration voice changes too — not just the words, the actual voice. Each register has its own distinct speaker with matching accent and affect. The default voice (no register selected) is a clean, neutral narrator.

Auto-play. With any register active, new responses narrate automatically after they finish streaming. You don't need to tap play every time — Asa just starts reading. Switch back to the default register and auto-play turns off.

Male and female voices. Toggle between male and female voice sets in the voice settings panel. Every register has both a male and a female default voice, and you can browse and preview alternatives for each one.

Custom voice picks. Don't like the default voice for a register? Open All settings → Voice to browse the full voice library. Preview any voice, and pick the one you want for each register. Your picks are saved per-register and per-gender — you can have one voice for Pirate male and a different one for Pirate female.

Streaming narration. Longer responses narrate in chunks as they load. You hear the first part while the rest is still being prepared — no waiting for the entire response before audio begins.

Looking things up

Every time you ask something, Asa looks through everything you've shared with him — your conversations, his earlier replies, your documents and your images — and answers from what he finds, not from guessing. You don't need to do anything special; it happens on every question.

He searches the web only when the question is about the world — news, prices, facts you haven't told him. So nothing personal goes out to the web unless it has to.

What Asa tells you a source said is linked to it. When his sources disagree — your notes say one thing, a website says another — he tells you so, with what each one says, instead of quietly picking one.

Gap awareness

Asa notices when you haven't been around. If you're away for a few days and come back, he knows time passed and picks up naturally — no awkward “how can I help you today?” restart. It's the difference between starting a new conversation and continuing an ongoing one.

Getting the whole answer

Ask for six chapters and you get six chapters, in one reply. Before a reply reaches you, Asa checks it against everything you asked for; if part is missing, he goes back and finishes it. You don't have to ask twice or tell him to keep going.

If something still can't be done — say you named four sources and only one is in your files — the reply ends by saying plainly what's missing, instead of passing off a partial answer as complete.

Sometimes a part needs something only you know, like how much your cat eats or which date you're moving out. Then the reply pauses on one question, with a box underneath for your answer. Answer it there and Asa finishes the same reply with your answer. If you'd rather move on, just type in the chat as usual.

Also uncovered. While making something for you or talking a thing through, Asa may notice something you didn't ask about but would want to know — a risk, a clash with something you said before, a connection between two things. He looks into it and adds what he found at the end of the reply, under “Also uncovered,” naming where it came from. When you ask a question of fact, the reply is the answer and nothing else.

Long replies aren't cut off at a fixed length. If you ask for something big, you get all of it.

Follow-up suggestions

After Asa finishes a response, you may see small tappable pills at the bottom of the message. These are follow-up prompts — things Asa thinks you might want to explore next, based on what the response actually said.

If the response cited a specific reference — a scripture verse, a paper, a person, an event — that reference becomes a pill. Beyond citations, Asa surfaces topical threads worth going deeper on. The pills are generated automatically; you don't need to ask for them.

Tapping a pill sends it as a new message, just like typing it yourself. Asa starts working on the follow-up immediately — you'll see thinking text that reflects what you tapped (e.g. “Exploring Romans 5:8”), and the response streams in as usual.

Not every response generates pills. If there's nothing meaningful to follow up on, none appear. They're there when they're useful, not when they're not.

Stopping a response

Changed your mind after sending? Tap the stop button in the compose bar while Asa is responding. The response stops immediately — whatever Asa had written so far stays visible, and you can send a new message right away. No waiting for a response you don't need.

Coding

Asa writes, reviews, and debugs code. Unlike a generic copilot, he carries project context from your conversation history — if you told him about your stack last week, he knows it this week. Paste code, ask for a review, describe what you want built. He works in whatever language you throw at him.

Multilingual

Asa responds in whatever language you write in. No configuration needed — write in Spanish, get Spanish back. Write in Hindi, get Hindi back. Switch languages mid-conversation and he follows. Your record is kept in the language you used.

The rest of Asa follows your language too: the chat's buttons and messages, settings, this guide, the terms and other pages, what's new, and the emails he sends you. Until you've written to him, that's your browser's language.

Fifteen languages can also be chosen by name: English, Arabic, Bengali, Chinese (Simplified), French, German, Hindi, Indonesian, Italian, Japanese, Korean, Portuguese (Brazil), Spanish, Swahili and Vietnamese. Pick one from the language menu on the home page, or as your register in the settings drawer, and Asa keeps to it whatever language you write in.

Asa never changes that choice on his own. If you ask him to keep to a language, a button appears under his reply, and the change happens only when you tap it.

Voice registers

Registers change how Asa speaks without changing what he says. Same information, different voice. There are nine registers, split into two groups:

Everyday registers — Southern Gentleman, Country Boy, New Yorker, Tenured Professor, Engineer, and Publication. Natural voices for different contexts — casual, regional, academic, technical, formal.

Theatrical registers — Pirate, Caveman, and Cyberpunk. Exactly what they sound like. Fun to try, surprisingly usable.

Every register has a male and a female voice. With a female voice, a few names change to match — Southern Lady, Country Girl, and Cavewoman.

To pick one, open the settings drawer (gear icon) and find the Register section. The default is Match my register, which means Asa mirrors your own tone — write casually, get casual back; write formally, get formal back.

Each register affects both the written text and the narration voice. When you switch registers, Asa's response is re-rendered in the new register's style and re-narrated with the matching voice — no need to send a new message. See Audio narration for details on voice playback.

Registers apply to English prose only. Code and tables are never affected. If you switch to a non-English language, the register is automatically suspended and restored when you return to English.

Think of it as a costume — the same content, told in a different voice. Technical accuracy is always preserved.

Connections — telling Asa about the people in your life

Asa can know about the real people in your life, but only if you tell him. This is what connections are for.

A connection starts with a vouch — you're naming what someone is to you. “Bruce is my father.” “Sarah is my friend.” “James is my coworker.” You pick from a set of relationship roles — immediate family, extended family (grandparents, aunts, uncles, cousins), friends, coworkers, mentors, neighbors, and more — so Asa always understands the relationship clearly.

For some roles, you can also add a modifier — “step” or “in-law.” These apply to the family roles where they make sense (father, mother, brother, sister, son, daughter, etc.). So you can say someone is your step-father, or your sister-in-law, without those being separate entries in a long list.

Once Asa knows about a relationship, he can reference it naturally. If you've told Asa that someone is your wife, and later you mention something about a conversation you had with her, Asa already has the context — he doesn't need you to explain who she is every time.

How it works

Sending a vouch. Go to All settings → Connections. Search for another Asa user by name, then:

1. Pick a role — tap the chip that describes what they are to you (father, wife, friend, coworker, etc.). This just selects it; nothing is sent yet.
2. Add a modifier (optional) — if the role you picked supports modifiers, a second row appears with “step” and “in-law.” Tap one if it applies, or skip this step.
3. Tap Connect — this is what actually sends the vouch. The other person will see it in their settings as “pending.”

Accepting a vouch. When someone vouches for you, it appears in your settings under “Connections about you.” You'll see what they called you — for example, “Bruce says you are their son.” To accept, you pick what they are to you from a matching set of roles. If Bruce called you his son, you'd pick “father” or “parent” — the options are constrained to roles that make sense together. Both sides independently confirm the relationship, which means Asa knows exactly how you relate to each other.

Withdrawing. Either person can end a connection at any time by withdrawing. This is final from your side — it removes both the vouch you sent and any acceptance you gave, in one action. The other person doesn't need to do anything; the connection is simply over. The history is still preserved — you can ask Asa about it — but it won't show up in your active connections anymore. If you want to reconnect later, you'd send a new vouch and the other person would need to accept it again. Just like real life.

Your settings

You can personalize a few things about how Asa works for you. All of these are in the settings drawer (gear icon) or under All settings.

Preferred name. This is what Asa calls you. By default it uses the name from your sign-in account, but you can change it to whatever you like — a nickname, a first name, whatever feels right.

Home location. You can set a general home location — type in a city or area and Asa uses it for weather, time of day, and local context. You can change it or turn it off whenever you want.

Location awareness. There's a toggle in All settings → Profile that enables per-turn geolocation. When it's on and your browser allows location access, Asa knows where you are each time you send a message — so he can notice things like “you're not at home right now” or reference what's nearby. You can turn it off in settings at any time, and you can also revoke browser-level permission separately. If location awareness is off, Asa falls back to your home location setting.

Appearance. Dark mode by default, or switch to light mode or let it match your device's system setting. There's also a groovy background toggle — on by default — that puts the swirling color from the landing page behind your conversation. All of this is in the settings drawer. Every setting you choose — appearance, voice, register, narration — is saved to your account, so it's the same on every device and stays put until you change it.

Switching to Extreme. Extreme is the most expensive mode, so Asa asks you to confirm before switching to it.

The red dot. A red dot on the gear icon means someone has sent you a connection request to review. You'll find it under All settings → Connections.

What's new. When Asa gains something new, a “New in Asa” pop-up tells you about it. The What's new page keeps the list of recent changes.

Sign out. At the bottom of the settings drawer. Signs you out of Asa on this device. Your data and conversation history are never affected — sign back in and everything is exactly where you left it.

Giving feedback

Every response Asa gives you has three small buttons — copy, thumbs up, and thumbs down. Copy puts the response on your clipboard. The thumbs aren't decorative — they go directly into how Asa learns.

A thumbs up tells Asa “that was good — do more of that.” A thumbs down tells him “that didn't work — something was off.” Each one is read with the reply it's on, quietly — rating a reply never starts a conversation.

Your ratings also help Asa beyond your own conversations: when ratings from several people point the same way about how he writes — too long for a quick question, say — he notes it about himself, and that note is reviewed before it's kept. No one's words go into it.

If something is more than just a “thumbs down” moment — if something is broken, confusing, or wrong — there's a Report a problem button right in the settings drawer (gear icon), or under All settings → Profile. That goes straight to the team behind Asa.

Privacy

Your conversations belong to you. Every user has their own completely separate record. Asa doesn't mix information between users — what you tell Asa stays in your space, and what someone else tells Asa stays in theirs.

The only place where two users' records touch is connections. When you vouch for someone (or they vouch for you), that creates a link between your records. But both people have to actively participate for that to happen — nobody can see your conversations or your information just because they vouched for you. For the full legal details, there's a Terms of Use link in the settings drawer.

Safety

Asa has a behavioral safety system that watches for concerning patterns across all your conversations — not keywords or topics, but what you're actually doing. Fixation on a person, escalating grievance, progression from thinking to planning — these are behavioral signals that accumulate over time.

If Asa detects a concerning pattern, he doesn't refuse your request. He engages with it — names what he's noticing, asks about the underlying intent, and steers toward constructive outcomes. If something reaches crisis level, he provides crisis resources including local services specific to your area.

The watching is done by a separate system, not by the AI writing your reply: it keeps track across conversations, tells Asa how to respond, and can send a reply back to be rewritten if it doesn't engage the way it should. If a conversation reaches a serious level, it's recorded and a person on the Asa team is emailed to look at it. For the full detail, see the Safety page.

Getting help

If you're stuck on something or something isn't working right, just ask Asa. He keeps his own notes on every feature — what it does, where the button is, its limits, and what to try when it misbehaves — and those notes are updated every time something about him changes. Ask “can I send you a PDF?” or “how do I change your voice?” and he answers from them, not from a guess.

He can also look at your own account — your balance, recent charges, and whether chat is paused — so “why did my balance drop?” gets a real answer. Nobody else's account, only yours.

When you need a person, tell him. He'll write up a note to the team for you and show it right under his reply. You can edit it, then press Send to the team — nothing is sent unless you do. The team replies to you by email.

You can also email the team directly at brandt@tryasa.ai.