The AI architect: an agent designed in one conversation#
It all starts with a sentence. You describe the task in the chat on the home page, the way you would explain it to a colleague: "Every Friday, send me the list of unpaid invoices with a draft reminder for each customer". The AI architect takes care of the rest:
- it asks its questions, only the ones that change the result: your sources, your schedule, your recipients, the tone to use, what it must never do;
- it asks for the connections it needs to your tools, in one click whenever possible;
- it shows you a sample of the deliverable, or a preview of the website for a project, before any payment;
- it quotes a fixed monthly price, based on how complex the task is.
You approve, and the agent is launched. If you start from a library template, the architect already knows the method and only asks for what is specific to you.
Two kinds of agents#
Automation agents carry out a task over and over: press reviews, follow-ups, reports, screening applications, payment reconciliation. They produce emails, files, rows in a table or drafts for you to approve.
Project agents build a hosted deliverable, most often a website. The agent builds the site, shows you a preview in the chat and applies your changes: the build and two revisions are free before you commit. Once launched, the site is hosted, can use your own domain name, and every month includes 10 revisions requested in a single sentence. If you want more, tiers with 100 revisions or unlimited revisions are available.
Agents that run on their own#
An agent can start in four ways:
| Trigger | Example |
|---|---|
| Schedule | Every weekday at 7:30 a.m., on the 1st of each month, every Monday at 9 a.m. |
| Event | A form submitted, an order received, an incoming message |
| Another agent | As soon as an upstream agent has successfully finished its work |
| On demand | You start the run from the dashboard or on WhatsApp |
Every run leaves a readable log: what the agent looked at, what it produced, and anything that got in its way.
More than 1,600 connected apps#
Agents work inside your tools, not next to them. Email, calendar, spreadsheets, CRM, accounting, e-commerce, project management, social media, file storage: more than 1,600 apps are available, with one-click connections. You grant access once, and the agent uses it on every run. The full list is on the connectors page.
Beyond apps, an agent can have built-in capabilities:
- web search and RSS feeds, for monitoring and gathering public information;
- file generation: PDF, Excel, Word, PowerPoint, CSV, images;
- code execution, for calculations and data transformation;
- sending emails from your own connected account;
- browsing websites that have no API;
- domain names: search and setup, with purchases only made with your explicit permission.
Deliverables ready to use#
An agent doesn't send you raw text to copy and paste. It produces the document you need: a formatted PDF report, an Excel workbook with its tabs and totals, a Word document to complete, a presentation, a CSV file for another piece of software. Files are sent by email or stored in the dashboard's Files area, where you can find them at any time.
Agents that work together#
Several agents in the same account can hand work to each other. A first agent fills a shared table (for example, the list of prospects found each day); a linked agent starts as soon as it has finished and only receives the new rows, to qualify them, write a message or update your CRM. This lets you break a complete process into simple steps, each handled by the best-suited agent.
A 3D map in the dashboard shows your agents at work and the links between them. The library offers ready-linked chains: KYC collection, checks and review, application screening and interviews, construction quotes, scheduling and progress claims.
Safeguards keep these chains in check: never after a test run, no loops, and a limited number of levels and triggers per day.
Individual and shared memory#
Each agent remembers. At the start of a run, it picks up what it already knows: your preferences, useful facts, summaries of its latest runs. While it works, it can save something important or search its memory.
Shared memory is common to all the agents in your account, and only to them: a piece of information noted once (your opening hours, your rates, the names of your contacts) serves your whole team of agents. No other account ever has access to it.
You can also give an agent context files: a price list, brand guidelines, a catalog, a procedure. PDF, Word, Excel, CSV, text, Markdown, JSON and images are accepted up to 25 MB; they are split and indexed so the agent can find the relevant passage. Just drag the file onto the agent's card.
A dashboard to keep track of everything#
The dashboard brings together your agents, their runs, their files, your projects and your subscription. For each agent, you can see when it ran, what it produced and how many times it has run this month. You can pause it, run it on demand or change its instructions. The Memory & files tab shows what it has remembered and the documents it consults.
WhatsApp: create and follow your agents from your phone#
The same architect answers on WhatsApp. You describe a task, answer its questions, connect a tool and launch the agent there, without opening a computer. You can link your web account with a code and receive your agents' results right in the conversation.
A library of templates#
More than 300 agent templates are ready to use: automations, websites and chains of linked agents, filterable by industry and by role. Each template describes its steps, its connectors, a sample deliverable and its monthly price. The use cases by industry present a selection for each line of business.
Clear pricing#
Each agent has a fixed monthly price, quoted before launch: from €3, with most agents costing between €9 and €49 per month. Three plans tie it all together:
| Plan | Price | What it includes | Storage |
|---|---|---|---|
| Free | €0 | You only pay for active agents | 1 GB |
| Pro | €29/month | 5 agents up to €9 included, 20% off additional ones | 25 GB |
| Studio | €79/month | 12 agents up to €19 included, 35% off additional ones | 200 GB |
Storage covers your files, shared tables, websites and your agents' memory. Details and a simulator are on the pricing page.
Safeguards by default#
- No real purchase (domain name, payment) without your explicit permission and a spending cap.
- No automated messages to strangers on social media: the agent prepares drafts for you to approve.
- Generated websites are served in an isolated environment.
- Daily limits apply to runs and to agent chains.