Help Center. How can we help?

DomaAI is a Polish AI assistant from MADD. It has two work modes: Ask and Computer, the latter with its own machine in the cloud. There is also Multi-Agent, a team of agents for harder tasks, 88 models from five providers (OpenAI, Anthropic, Google, xAI, Mistral AI), image and video generation, a calendar in the chat, Office documents and your own AI agent with a name, look and character. Native in Polish, on iPhone, iPad and in the browser.

MADD is a Polish company (MADD Marcin Kisieliński, Olsztyn). We build DomaAI in Poland with the Polish user in mind: the interface, agents and documents are designed in Polish, not translated from English. We keep the account, chat history and files on servers in the European Union (OVHcloud).

DomaAI is on the App Store for iPhone and iPad with iOS 26 or later. Updates install through the App Store. You can also use the chat in the browser at domaai.pl/czat, with no installation, on the same account.

iOS 26 or later and an internet connection (Wi-Fi or LTE). Image and video generation runs on the server, so it does not load the device. In the browser, any current Chrome, Safari, Firefox or Edge on Windows, macOS and Linux works.

Yes. The Free plan is free and needs no card. New accounts created by 31 October 2026 get a one-time 1000-credit welcome bonus, enough to test Multi-Agent, chat, images and most features. Higher plans (from 39.99 PLN per month, and on the website with 10% more credits) you unlock only when you need more.

After the first launch:

1. Create an account by email, with Apple ID or Google (a dozen seconds or so).
2. Write your first question. The assistant picks the model itself, and in Multi-Agent mode it brings in a team of agents.
3. Try a complex request, e.g. “plan a weekend trip to Kraków and make a budget”.
4. On iPhone and iPad (from version 0.1.19), with your first message you set up your own agent: name, look, character, style and brain. Later you can change this in Settings, in the “Your agent” section.

In the browser you don't need to configure anything: the conversation works from the first message.

The interface has 11 languages: Polish, English, German, French, Italian, Dutch, Czech, Russian, Ukrainian, Japanese and Korean. You choose the language when logging in or in settings. The assistant talks in each of them, and in Polish it writes naturally, not as a calque from English.

In a regular chat one model answers. Multi-Agent gives a harder task a team: from a single agent (solo), through teams with leads for research, analysis, code, texts, planning, calendar and images, up to Deep Research with multiple iterations. You see the scenario and its cost before the start, and you get the result as a finished document, not a wall of text.

Yes. Multi-Agent works in all plans, including Free. The differences are in run limits (Free 15, Base 30, Plus 80, Pro 200, Max 500 in a 24 h window), in team size and in access to top models. A run costs from 5 credits (solo), through 20-55 credits (team) and 140 credits (Deep), up to 320 credits (Deep Research), plus the tokens of the models used.

Web and news search from the last hours, reading web pages, calculator, weather, search of your chat history, analysis of PDF, DOCX and XLSX documents, image generation, calendar and project memory. The agent decides for itself when to use them, and you see each use in the line of thinking.

Yes. While the team works you see live which agent is active and what it is doing: who is searching the web, who is calculating, who is verifying. The thinking panel shows the next steps, rejected sources and the reasons. You can check the answer, not just take it on faith.

A simple question takes a few seconds, as in a regular chat. A team task takes longer, depending on the scenario and the number of searches, and Deep Research takes the longest, as it compiles a report from many sources. You can leave long tasks in the background and close the app; progress shows on the lock screen, and the result waits in the conversation.

Each project has its own memory: decisions, files and context stay between conversations, so you don't repeat the introduction. You manage the memory yourself: you can review, edit and clear it in the project settings. The capacity of long-term memory depends on the plan: from 50K tokens (Base) to 1M (Max).

It is the second work mode of the assistant, next to the default Ask. You switch it in the “Ask | Computer” island above the text field. The assistant then gets its own computer in the cloud (Ubuntu) for the duration of the conversation: it installs what it needs, builds, runs, checks the result itself and fixes it until it works. You see the task plan with checkable steps, action cards (commands, changed files, tools used) and a timeline. Computer mode is available from the Plus plan.

Ask, when you want an answer, a document, an analysis or one specific matter to handle. You get the result in a dozen or so seconds, with nothing to install. Computer, when you commission a project: a website, a tool, reworking hundreds of files, something that has to work afterwards. It takes minutes to hours, but you can put the phone down and come back for the result. A comparison of both modes is on the home page.

Company websites, landing pages, internal tools (a pricing calculator, an offer generator, a panel with data), clickable prototypes in the browser and code packages to hand over to a team. A finished site is published at name.domaai.pl; the site card in the conversation shows its status live. The site panel has the tabs Preview, Code, Panel, Database, Files, SEO (score 0-100 and fixes with one tap) and Settings with your own domain and DNS instructions. You add modules with one tap: Stripe payments, shop, bookings, login, customer panel, reviews, form, newsletter, AI chat, blog, gallery, map, analytics.

Yes. Tasks run in the background: leave the app, and when it finishes a “Done” notification arrives. The machine lives for the whole conversation, so the next day you return to a ready workshop: the project, data and version history are waiting in place. You hand repetitive work to an automation agent - see the section Agents and automations.

Computer mode is included in the Plus plan and higher. Individual actions (model messages, search, images) are billed in credits just like in the chat, and you see the cost in the task run. Publishing a site at name.domaai.pl is hosting for a set time (from 15 minutes to 30 days) or with your own machine. It is billed in credits, you see the price before publishing, and you can extend, withdraw and republish the site. The Hosting tab is available from the Plus plan.

No. Irreversible effects - a purchase, sending to a third person, deleting data - require your confirmation in the conversation, and automation agents work according to the permissions you give them (Auto, Ask, Never). Before a risky change the assistant saves a checkpoint, which you can return to with one move. You can watch the browser window of the agent and take it over at any time, and it saves a successful path through a portal as a recipe only if it contains no irreversible steps.

It is your AI worker with a name, role, instructions, model and permissions. You attach it to a task and it carries it out on its own - on a schedule or by webhook, also when the app is closed. This is different from the Multi-Agent team, which works on one message in the chat. You will find an overview of the capabilities on the Automation agents page.

You create an agent in the conversation (on iPhone, iPad and at domaai.pl/czat) or in the app on the Automations screen.

• In the conversation write what the agent is to do and when, e.g. “Create an agent: check news from my competitors every day and let me know in the morning”. The assistant will pick a template or suggest its own agent and ask about any missing details.
• In the app: “+” menu → Agents and automations → Agents tab → New agent. Choose one of 8 roles - the instructions fill in by themselves - and set the permissions and the credit limit per run.
• In the Tasks tab tap “New task” and choose the agent.

On the Automations screen you open templates from the ⋯ menu in the top right corner, with the “Templates” button on the Agents tab, or via “New task” → “From template”. A template sets up the agent together with the task, and you fill in a field or two. In the conversation it is enough to ask for a template by name, e.g. “Create an agent: morning mail review”. The full list is on the Automation agents page.

• Auto - the agent does it itself, and the action goes into the history.
• Ask - it prepares the action and waits for your consent.
• Never - no access; it does not even see the tool.

By default, reading (mail, calendar, documents) is on Auto. Everything that goes outside - an email to another person, a WhatsApp message, publishing a site - is on Ask. Turning on spending of the ad budget and creating further tasks are blocked by default. Ad budget spending cannot be set to Auto, and Auto for a high-risk skill requires separate confirmation.

You get a notification that the agent asks for consent. In the Consents tab, or on the card in the chat, you see a preview of the action and choose Approve or Reject. You approve exactly what you see: one consent covers one specific action. When a request waits 6 hours, we remind you with a single notification, and without a response it expires after 7 days. You can also decide in the browser - the link in the result email opens it in domaai.pl/czat - or on WhatsApp, if you have linked your number (see the section WhatsApp).

You give the command, the agent and the rhythm: daily (also only on business days), weekly, every N hours (from every 24 h on Free to every hour on Max) or a webhook (from the Base plan). You get the result by email and, if you want, also as a push. A task created in the conversation can also return its result to that conversation - just tell the assistant. You can also run a task right away with the “Run now” button.

An address to which a form on your website or another service sends data (POST with JSON). Each call starts the task, and the data goes into the agent's command. You get the webhook address after saving the task - treat it like a password.

As much as the agent's work: model messages and tools are billed in credits, just like in the chat. A run has a starting cost of 20 credits (55 credits for tasks on the Computer machine), and each agent has a limit per run (default 60 credits, maximum 500 credits). Once it is used up, it ends the work and describes what it managed to do. It never spends more than you have in your account. You see the cost of each run in the task history.

It depends on the plan: Free - 1 agent and 2 active tasks (usually every 24 h, no webhooks), Base - 3 agents and 5 tasks (every 12 h), Plus - 10 agents and 15 tasks (every 6 h), Pro - 25 agents and 40 tasks (every 3 h), Max - 50 agents and 100 tasks (every hour), Enterprise - no limit. Agents work in every plan, including Free (on the credits from the welcome bonus or from a pack). After you switch to a lower plan, the enabled tasks keep running at their rhythm, but the daily run limit of the new plan applies immediately. Webhooks and runs per day are in the limits table.

The internet and browser, Gmail, Google Calendar, GitHub, WhatsApp, Office and PDF files, and code in a sandbox. You connect a service in Connectors; for a permission that requires it you will see a “Connect” button. Meta ads require Meta's consent: the Meta Ads row is visible in Connectors with the label “Requires Meta consent”, but connecting the account is temporarily unavailable, because Meta has not yet granted us permissions for ads and pages. Until then the assistant does not create or run campaigns and does not publish on Facebook or Instagram.

In the task details, on the History tab: the result, the cost and the action log of each run. We keep the run history for 180 days (the last 20 runs of a task are kept longer), and the action log for 90 days. You can also ask the assistant about the last result in the conversation.

Turn off the task with its switch, turn off “Agent active” or delete the agent. If the agent has active tasks, the app will ask whether to turn them off too. You can stop all tasks at once with the command “Pause all tasks” from the ⋯ menu on the Automations screen. In the conversation it is enough to write, e.g. “pause the morning mail review”.

When three runs in a row fail (e.g. an error or no credits), we pause the task ourselves, and you see the reason in its details. You turn it back on with the same switch.

Yes. In domaai.pl/czat you create an agent in the conversation, decide on consents (Agent consents in the side panel or the link from the email), and turn tasks on and off with the switch in the “Routines” section of the window that opens from the “Machines” tile in the side panel. Detailed editing of the agent (role, instructions, permissions, limit) is in the iPhone and iPad app.

88 models from five providers: OpenAI (GPT-6 Astra, GPT-5.6 Sol, Terra and Luna, GPT-5.5, o3, mini and nano), Anthropic (Claude Fable 5.1, Fable 5, Opus 5, Sonnet 5, Haiku 4.5), Google (Gemini 3.7 Flash, 3.1 Pro), xAI (Grok 4.6) and Mistral AI (Large, Medium, Codestral, Devstral). Each is connected directly via the producer's API, with no middlemen. New models arrive on an ongoing basis, and you will find the full list with rates in the credit pricing.

They are switched off until the safeguards for transferring data outside the European Economic Area are formalized (Art. 46 GDPR). No user data is passed to them. If they return, it will only be after standard contractual clauses and a data processing agreement are concluded, and then we will update the subprocessor list.

Standard models are fast and cheap (e.g. GPT-5.6 Luna, Gemini 3.7 Flash, Claude Haiku 4.5, Grok 4.6), available at a credit rate from Free or Base. Top models are flagships (GPT-6 Astra, Claude Fable 5.1, Opus 5, Sonnet 5, GPT-5.6 Sol): not available on Free and Base, on Plus 50 messages in a 24 h window, and from Pro with no limit. You see the rate of each model before sending a message.

You don't need to know the models - the assistant picks them itself. If you want to choose manually:

• Writing and conversation - Claude Sonnet 5 or GPT-5.6 Sol
• Code and difficult analyses - Claude Fable 5.1, Opus 5 or GPT-6 Astra
• Fast, cheap answers - GPT-5.6 Luna, Gemini 3.7 Flash, Claude Haiku 4.5
• Long documents - Claude Fable 5.1 (a million tokens) or GPT-6 Astra (922K)
• Processing in the European Union - Mistral AI models (France)
• Research with internet access - Multi-Agent with search, or longer thinking

Yes, at any time - the model choice is above the message field. The conversation context stays, so you can start with a fast model and switch to a flagship for the harder part. You can also regenerate an answer with a different model and compare the variants.

Yes. We use the APIs of OpenAI, Anthropic, Google, xAI and Mistral directly - these are the same models as in the producers' apps. No clones or local distillations. Requests go to providers with training on client data disabled.

Nano Banana (Google, in the Lite, 2 and Pro variants) and Grok Imagine (xAI). Generation from a description, editing your own photos with a command in Polish (“add rain”, “change the background”), a chain of corrections in one conversation and style transfer. Images are billed in credits (from 30 credits per image), and the daily limit depends on the plan: from 1 image on Free to 150 on Max.

Veo 3.1 (Google, Lite, Fast and full variants), Sora 2 and Sora 2 Pro (OpenAI), and Grok Imagine Video (xAI). Video from a description and from an image (bringing photos to life), some models with native sound. The rate is counted per second of video and visible before the start. Video generation works from the Base plan, with a monthly limit: Base 2, Plus 5, Pro 15, Max 40.

You draw the path of movement on a photo with your finger, and the assistant turns it into a short film with that movement. The feature is in the Video tab on iOS from 24 August 2026.

Image from 30 credits (Grok Imagine), through 52 (Nano Banana 2 Lite), up to 206 (Nano Banana Pro). Video from 75 credits per second (Veo 3.1 Lite, Grok Imagine Video), up to 600 (Veo 3.1). Exact rates are in the credit pricing and in the app before each action.

Depending on the model and length - usually from a dozen seconds to a few minutes. Generation runs on the server, so you can keep writing or close the app in the meantime. You see the progress in percent in the conversation and on the lock screen (Live Activities), and the finished clip waits in the library.

Yes - content generated in DomaAI you may use commercially, in line with the terms of the producers of individual models. Do not generate material that infringes others' rights, trademarks or likeness. Details in the Terms and the Usage rules.

Live voice conversation is temporarily switched off. You can dictate a message with the microphone on the iPhone keyboard, and listen to the answer with the “Read aloud” button under the answer.

When you ask for an email, a recipe or a document, the answer comes as a card ready to use: the email opens in your mail client with the ready content, you save the recipe with one tap, and you download the document as a Word, Excel or PDF file. You can edit the cards before using them.

It creates PPTX presentations, DOCX documents, XLSX spreadsheets with charts and PDF files, built in the chat and ready to download. It reads PDF, DOCX, XLSX and photos: it summarizes, extracts tables and points out risks in contracts. There are also HTML artifacts, code with a preview, and Canvas, where you edit text together with the AI.

Yes - with Apple Calendar and Reminders. Say “team meeting on Friday at 12:00”, and DomaAI detects the date, checks conflicts with what you already have and saves the event after your confirmation. It adds nothing without your consent.

The Whisper feature joins a meeting on Zoom, Microsoft Teams or Discord, listens to the conversation and creates a live transcript, and after the meeting a summary and a list of tasks. No recording and no plugins on the participants' side. You pay in credits for the transcription time.

You share each conversation with a link from the conversation menu. The link opens in the app or in the browser, including for people without an account. You switch sharing on deliberately with a button, and you can turn it off at any time.

Yes. Group chats let you add friends or a team to one conversation with the assistant - everyone writes, and the answers and generated content are visible to all. You get notifications about new messages on your phone.

The number +48 690 746 669 runs the company bots that use DomaAI. When you link your own phone number to the account (in the app: WhatsApp → “Connect number”, then you send a code on WhatsApp), you also settle agent consents there: write “zgody” (consents) and the number will show the waiting requests with codes. Reply “OK” or “Odrzuć” (reject) with the code from the list. You approve an important action only in the app with Face ID, and on WhatsApp you can reject it.

The number does not send consent requests or agent results by itself: you get those in the app (push notification) and by email.

The conversation with the assistant is in the DomaAI app on iPhone and iPad and at domaai.pl/czat. The DomaAI WhatsApp number serves company bots and answering agent consents. For questions to the assistant, go to the app.

You create the bot in the app on iPhone and iPad or in the browser (domaai.pl/czat): name, description, instructions and AI model. The bot answers your customers on the DomaAI number, and the conversation opens with your “Start @name” link. The customer writes “menu” to end the conversation with the bot.

• At the first reply the bot says it is an AI assistant, and gives the link to the information on data.
• “człowiek” (human): the bot passes the request on and goes silent in that conversation for 24 hours; before that you can resume it in DomaAI in the browser (domaai.pl/czat). You get a notification and an email, and you reply to the customer from the app within 24 hours of their last message.
• “STOP” - the bot stops writing to that person, “START” resumes it.

Replies are billed from the credits of the bot's owner, according to the AI model chosen for the bot. From 1 October 2026 Meta also charges a fee for each message sent on WhatsApp. We add it to the credits and show it in the credit history. When the credits run out, the bot stops answering, and you get an email.

Coming soon. We are preparing the connection of your own company number via Facebook, including a number you use in the WhatsApp Business app, without giving that app up. Until it launches, the bot answers on the DomaAI number. We will let you know in the app when the connection is available.

Your own number requires a company account on Facebook and a card added in Meta, because Meta bills the messages from the company's number directly to the company.

The controller of the data is the company that runs the bot. MADD processes it solely on its behalf and does not use it to train models or for advertising. Details, including in English: Information for people in conversation.

Yes, at any time - in the account settings in the Subscription section, or in the App Store if you bought it through the phone. The paid period works until its end. After expiry the account returns automatically to the Free plan without losing history.

Credits from purchased packs do not expire as long as you have an account, and they stay on it even after the subscription is cancelled. Credits from a subscription are valid until the end of the paid period.

Yes. A VAT invoice is issued automatically after each payment on domaai.pl and is sent to KSeF. You provide your company details (NIP) before purchase, and you download invoices in the account settings in the Payments section. For App Store purchases, Apple settles the payment and Apple issues the document.

You buy subscriptions and credit packs on the website (by card via Stripe) or in the iOS app through the App Store - by card, Apple Pay or Apple balance. The Enterprise plan can be paid by bank transfer against an invoice. All prices are gross and include 23% VAT.

Yes. When moving to a higher plan you pay the difference pro rata to the end of the period, and you immediately get the new limits. When moving to a lower plan the change takes effect from the next billing period - until then you use the current one.

Yes. In settings: Account, Delete account. With Apple or Google sign-in the app will ask you to confirm your identity again. We delete all data within 30 days in line with GDPR. The operation is irreversible.

No. DomaAI does not use your conversations to train models. Training is disabled at all providers (OpenAI, Anthropic, Google, xAI, Mistral AI), and we have data processing agreements with OpenAI Ireland and Anthropic Ireland. Details in the data policy and model training.

Yes. In transit via TLS 1.3, and data at rest sits on an encrypted volume (LUKS, AES-XTS). Backups are encrypted with AES-256. Tokens of connected services, for example Gmail or GitHub, are encrypted and never reach the models.

We store the account, chat history and files on servers in the European Union (OVHcloud). Requests to AI models go directly to their producers: Mistral AI processes them in the EU, the other providers in the US. The subprocessor list describes the country and transfer basis for each provider.

Yes, in line with GDPR you have the right to a copy of your data. In account settings you download the data export as a ZIP file, and there you can also freeze or delete the account. If you have questions, write to prywatnosc@madd.im.

Use the form on the Reports page or e-mail bezpieczenstwo@madd.im. We handle reports in line with the publication policy and the Terms.

Quick checklist:

1. Check your internet connection (switch between Wi-Fi and LTE).
2. Close and restart the app.
3. Make sure you have the latest version from the App Store.
4. Restart your phone if the problem keeps recurring.

If this doesn't help, write to pomoc@madd.im with your phone model and iOS version - we will help.

Check that you are logging in with the same method you used to create the account (email, Apple ID or Google). When logging in by email, use the “Forgot password” option. If the verification code doesn't arrive, check your spam folder and wait a minute. Still not working? Write to pomoc@madd.im.

On a weaker connection (e.g. LTE on the move) the answer stream can break off. The app fetches the missing part by itself after a moment; if it doesn't, refresh the conversation by pulling down. Credits for a failed answer come back automatically.

The most common reason is not enough credits - check the balance and the cost of the chosen model before starting. Also make sure the description does not break the content rules (models reject, for example, likenesses of real people without consent). If the provider has an error, try again in a moment or choose another model.

The Meta Ads connector is labeled “Requires Meta consent”. Until Meta grants DomaAI permissions for ads and pages, connecting the account is unavailable - this is not an error on your side. Until then the agent cannot read, create or change campaigns. Once Meta gives its consent, the connector unlocks in Connectors.

You report bugs and ideas right from the app (Settings, Report a problem) or by email to pomoc@madd.im. We reply within 24 hours on business days. The more details (what you were doing, what happened, a screenshot), the faster we fix it.