Transcript
Translated into English:
Hello everyone, welcome. I used to introduce myself as a growth marketer in startups, but today I consider myself a builder in the age of AI and no-code. My background spans a carbon accounting startup, building tools at the Ministry of Ecological Transition, freelancing, and launching side projects. Seven months ago, I joined Softr full-time, working in growth marketing while continuing to build.
Historically, Softr has been known as an accessible, beginner-friendly tool that remains flexible. Six months ago, we looked at bypassing limitations by triggering webhooks to external tools like n8n, but the platform has evolved significantly since then. Today, you can go from a single prompt to a functional application using AI, removing much of the manual setup barrier.
At the same time, Softr retains traditional visual no-code development, where you manually choose and configure blocks. For about a year now, we have also had Softr Databases built directly into the platform, similar to Airtable. This allows you to manage your data source natively alongside your front-end interface, while still supporting connections to external databases.
With the addition of Softr Workflows late last year, the platform now covers the complete stack: data, front-end interface, and logic. Previously, builders relied on a combination like Airtable for data, Softr for the portal, and Make for automations. Now, you can build your entire application natively within Softr using visual standards across databases and workflows.
We see this hybrid approach emerging in standard no-code tools like Airtable or Make, combining classic building with AI. AI makes no-code tools even more accessible, which is really interesting. Let us share the screen and dive into the subject.
Here we are on the Softr dashboard. I am very happy to have joined the Softr team because it is one of the tools I use the most daily, alongside n8n, notably for my SaaS Lookio. Now I can build directly with the team, bring feedback, and help make product decisions as a builder. For instance, we already plan to display recent applications directly on the homepage.
We know training organizations have to constantly update their courses because the platform evolves quickly. We are improving how we communicate major updates. Now on the homepage, there is a prompt box to get started with the AI Co-Builder, which launched about a month ago and helps during both creation and editing. Today, you can start from a blank page, from a template, or by entering a prompt.
Regarding Softr's vision, co-founders Artur and Mariam explained it well in a recent video: the core goal has not changed, which is to allow non-technical people to build business software. Historically, you built everything manually. Now AI provides a second option to assemble applications using the same underlying components.
It is entirely your choice to start from a template, from scratch if you want full control, or via AI. We offer this hybrid approach to make the process as easy as possible. Previously, many users would start in Softr but struggle to reach a complete, functioning application.
There is a barrier to entry where you need to understand quite a few things and figure out how to lay out your interface. Unfortunately, some people give up before they even hit publish. With AI, you can reach the point where you have something ready to publish and test. I learned a lot using the AI Co-Builder because it uses layout patterns that I find interesting. Often, I use the AI Co-Builder to create a new app initially, and then I continue building mostly by hand.
I will pitch this project live and use Whisper to transcribe the prompt. I want an application for my team to organize online webinar events. We will have admins who can do everything, including adding users, and managers who manage the events. We need a webinars database and a speakers database, with a bio field for each speaker that we will use later for workflows. Users can create an event, link speakers to it, and fill in descriptive fields.
I am using Whisper Flow for voice transcription. Dictating lets you provide more detail quickly, and the more behavioral context you describe, the better the AI results will be. The tool has now generated the requirements step. It asks context-specific follow-up questions, such as clarifying status options like draft, published, and completed, or confirming that a webinar can have multiple speakers.
It also proposes interface pages like webinars, speakers, and a dashboard. The AI Co-Builder usually limits itself to between zero and three clarifying questions, or moves straight ahead if your initial prompt is detailed enough. Next is authentication. Softr handles all standard business app logic by default, so you do not need to vibe-code everyday essentials.
For most internal apps, you do not want an open sign-up page because only an admin should add teammates. Softr also handles standard requirements like 404 pages and password-reset flows automatically. We can set navigation to a sidebar layout, which works well for internal tools, and then select one of about twenty available themes before generating.
C'est une préférence : j'aime beaucoup Zen et je l'ai réutilisé pour un futur projet. À droite, ça se met à jour avec une petite preview. Certains styles sont différents ; il y en avait un qui ressemblait beaucoup à LinkedIn, mais ce flow de démarrage est intéressant et tu peux ensuite changer les couleurs.
L'idée, c'est de gagner du temps. On avait une option qui allait chercher ta marque à partir du nom de domaine de ton email, mais elle est désactivée ; on verra comment retravailler ça. Pour l'instant, on part sur Zen et j'envoie « Create Your App ».
On reçoit un brief ici. J'ai demandé qu'on puisse davantage le revoir, parce qu'au final, c'est plus compliqué de tout modifier et déconstruire ensuite, même si on entraîne notre AI. Je suis dans un workspace particulier, Templates, donc la base de données porte ce nom ; je crée peut-être cette app pour en faire un template officiel Softr.
On voit que la base de données, le point de départ d'une app, est construite avant même de créer l'interface. La donnée est le cœur, et beaucoup de gens commencent par l'interface en oubliant de travailler les données. Là, il a créé le schéma, modélisé les intervenants, puis rempli la base avec des données de test.
C'est hyper important : quand tu démarres un projet sans données, c'est compliqué, donc il faut vite en injecter. Les devs utilisent des librairies de fake data, comme Faker, mais à l'ère de l'AI, c'est beaucoup plus simple. En live, les pages, boutons et blocs se sont ajoutés sous nos yeux.
Ce sont des blocs Softr. Si tu demandes quelque chose de très custom, il peut éventuellement ajouter un Vibe-Coding block, mais ici il utilise les composants natifs de Softr. C'est important, parce que tout ce qui a été créé reste modifiable visuellement.
C'est comme voir quelqu'un construire très vite le layout, et ça permet aussi d'apprendre à utiliser Softr. Il explique ce qu'il a fait, et tu peux même ouvrir les commandes lancées. L'intérêt, c'est de créer un outil business sans coder chaque ligne ni gérer 120 fichiers, dépendances et autres éléments d'un projet 100 % codé.
On compose une app avec des blocs préexistants : pas besoin de les réinventer, l'authentification, la base de données et le code sont gérés. Des AI génèrent des JSON qui définissent une page, ses blocs et leurs configurations ; on ne code pas, on agence et paramètre.
Il a créé la base de données et les interfaces : crée-t-il aussi les workflows, la logique back-end ? Il existe un AI Co-Builder pour Softr Workflows, mais il n'intervient pas encore dans ce flow de création d'app. On travaille déjà à lancer ce flow depuis une base existante, Notion, HubSpot ou une Softr Databases existante, au lieu de créer une nouvelle Softr Databases.
Right now, this only works with a Softr database. You cannot point it to an existing database yet, though the team is already working toward that goal. This first version creates an app from scratch, and it does not integrate Softr Workflows immediately, but that will come later. It is useful to confirm the database structure first before adding logic.
Could we provide a full design system in Markdown prompt instructions to customize the look? There are two aspects to consider. Currently, the flow requires picking a theme, but if a prompt already specifies design instructions, we could skip that step and propose a matching theme. However, Softr is not designed for pixel-by-pixel styling like custom block transparency or double padding.
Softr focuses on business apps, internal tools, and client portals rather than public consumer software competing with custom web stacks. Our lead designer, Austin, intentionally limits granular cosmetic options so users cannot create flawed UI or UX patterns. Users expect components like a kanban, form, or calendar to behave predictably right out of the box.
For internal business applications, functional UX matters far more than endless visual tweaking. In older versions of Softr before 2025, people spent hours restyling raw labels and values because default blocks were unpolished. The new blocks introduced in January 2025 are modern by default, saving substantial configuration time while leaving room for deeper customization through the Vibe-Coding block. Let us click Try it live to see what was generated.
We have the option to preview, which is super powerful because we can preview this app as a specific user, notably an admin here, and switch over to see the manager view. Our app lists webinars, includes a filter, and provides a search bar. We can click on a webinar to see more details and check which speakers are linked. Here, we also have our speaker database.
There are a few design details I gave feedback on. For instance, it is not immediately obvious that an element is clickable to view the record detail page, so I prefer using an item button labeled Open. Overall, from a single prompt, we have a kanban pipeline for webinars where we can drag items with real data to test the workflow immediately.
It works like a CRUD system, handling database operations while presenting and manipulating data clearly. A business app serves a purpose, usually managing and tracking things. Here, it manages events and a speaker database, letting us associate them easily.
We set up user groups like admin, manager, clients, students, or teachers, each with specific permissions and page visibility. There is also a feature I really like called Ask AI. It is natively enabled with a toggle, adding an AI chat that answers questions based on the data inside a block.
Clicking Edit returns us to Softr Studio, where we can close the chat and configure blocks visually in no-code. We see the generated pages, which we internally call entity pages, representing lists of speakers or webinars. Clicking a record opens an item details block on a detail page, which you design once like a CMS template for all matching items.
On the speaker detail page, we show the photo, name, title, and LinkedIn URL. Below that, another section lists all webinars linked to that speaker using a filter. If you want to modify a specific block conversationally, you can open the AI Co-Builder at any time.
The AI Co-Builder knows our context based on the selected page and block. We can ask it to add a CSV export option to download a speaker's list of webinars. We aim for feature parity between what you can configure manually and what the AI Co-Builder can do.
The AI added the export button directly to the block settings. We can easily rename the button label to Export CSV manually rather than spending AI credits on minor tweaks. I also frequently prompt for an analytics dashboard page, which we can add next.
We might just want to see the number of webinars per day or the type of webinar. We can simply prompt it to add an analytics page, and it will add dashboards and charts. That is how this hybrid experience works.
Regarding the difference between Softr, Lovable, Bolt, and tools like Claude Code, it comes down to the development approach. Tools like Lovable write actual lines of code, create files, install dependencies, and build an application from scratch. Softr is an established app builder with pre-built blocks dedicated to business applications.
In Softr, AI arranges these existing blocks, configures them, and connects them to a native cloud database. The databases use a visual no-code approach. We also have a built-in user management system tailored for business apps, where we can define user groups like admin or manager with conditions to handle essential permissions.
With Lovable, you control every pixel and have total freedom, but you must ensure everything works, from workflows to database security and RLS policies. Softr offers a structured framework that provides more security and stability out of the box, with options to customize further.
With something like Claude Code, you have to manage hosting, local execution, and database administration like Supabase on your own. Every project requires reinventing the wheel. A local SQL database gives you nothing visual to administer it, whereas Softr provides a native visual database interface.
In Softr Databases, you can transform a column into an AI field that searches the web to write descriptions or enrich company data automatically. It raises the question of whether it makes sense to launch a full-stack, fully coded project just to replace a spreadsheet. Instead of setting up Cloudflare, Supabase, and Stripe, you can use a dedicated tool designed as the evolution of a spreadsheet for internal tools or client portals.
Tool choice depends on technical skills and project needs. Not everyone can use Claude Code, despite the hype. Many projects simply do not require that level of low-level flexibility.
In a professional context, business apps usually serve the same core purpose: administering, visualizing, and leveraging a database. Someone in the chat asked to see Softr Workflows alongside the AI Co-Builder, and I would also like to see the Vibe-Coding block. Additionally, it would be interesting to see how you display charts, like webinars per person.
We have always had charts, and following a redesign, we have a few options available here. They are relatively simple right now and still improving, but you can quickly set up a chart to count webinars by month with a clear title. You can also configure sub-grouping, for instance by status or category, to get quick visual breakdowns.
Another major update is the ability to build richer page layouts instead of purely vertical pages. I built a CRM template where almost everything lives on the homepage alongside a few tables, displaying deals, contacts, and companies directly. With multi-column containers, such as two or three columns, you avoid navigating to separate entity pages for every single item.
While you can vibe-code charts, I recommend relying on native components for at least 80% of your business app since there is no need to reinvent standard elements. Having the Vibe-Coding block is still essential because pushing past standard tool limitations is often a concern when evaluating builders. A great native example is the calendar block, which is very practical.
You connect the calendar block directly to the webinars table, map the title field, set the start time to the date field, and enable day, week, or agenda views. With three clicks, clicking an event opens an item details block or page where users can view and edit records. There is no reason to vibe-code a standard calendar when the native one works this well.
However, you might vibe-code custom features, like a tool I built that took a meeting transcript, used AI to break it into subtasks as database records, and displayed them beside a calendar for drag-and-drop scheduling. For highly custom interactions like that drag-and-drop interface, you use the Vibe-Coding block. We can demonstrate this by combining the block with Softr Workflows.
We can build an AI helper that recommends speakers based on their bios, which acts as the context data the AI evaluates. This pattern applies broadly, such as building a product recommendation engine. We will create a form to input an event title and description, click a button, and have AI return the top three recommended speakers using Softr Workflows and the Vibe-Coding block.
I will open a new table and start creating the workflow. In business contexts, a workflow triggers when you have a new support request, and its actions categorize it, assign it to the right person, or notify a Slack channel. In our case, I will use a webhook trigger to communicate with the Vibe-Coding block. I need the webhook URL, which acts like a phone number that the Vibe-Coding block calls to interact with the workflow.
You might wonder if we could do this without a workflow. We could have the block create a record that triggers a workflow, or attach an action to each item where clicking it runs a workflow. We are also exploring AI directly integrated into blocks for lightweight local tasks like grammar correction or voice memo transcription sent straight to your database.
[.blog-callout]
Note from Softr: Softr Workflows let you trigger automated logic via webhooks, record updates, or custom button actions directly within your application.
[.blog-callout]
Right now, communication goes through webhooks, but we plan to build more native connections so a block can trigger or even compose a visual workflow attached to it. A key aspect is that you can connect natively and securely to Softr Databases, any API, or public datasets using custom credentials or OAuth. Even though our database queries will happen inside the workflow itself, blocks can natively read, create, or update records.
Now I will write a prompt for the Vibe-Coding block. I want an interface where I can write an event brief for a webinar, click a button, and receive a suggestion for the top three speakers. The block will send the brief and the user record ID via a POST request to our webhook URL, then listen for the response containing the recommendations. We need to define the expected response schema, which we can easily draft with ChatGPT based on our database structure.
We had the speakers' names and their LinkedIn URLs, for example. I'll prompt: write me a very simple JSON array listing three speakers, with a key for their name and a key for their LinkedIn URL for each speaker, and place that inside an object called speakers. I am anticipating what the response from this webhook will look like so we know how to display it in the interface. We will have a field called speakers containing a list of three items, each with a name and URL.
While it generates, someone asked about the AI models used to create blocks and whether we have a choice. By default, we have a balanced model that works very well. For small edits, there is a lighter, faster mode that uses fewer AI credits. You cannot use your own API keys because everything runs through Softr using AI credits.
The balanced mode runs on an Anthropic model. At Softr, our role with the AI Co-Builder is simply making sure it works seamlessly alongside cloud databases and user permissions. However, in Softr Workflows or Softr Databases AI, you can choose your specific model because you are building custom workflows for production. Abstracting model choices into simple options keeps things straightforward for non-technical builders.
Looking at the block configuration, under content, we have a no-code interface to modify settings directly. We also have access to the underlying React code, which you can copy into tools like Cursor or Claude Code to iterate on. We provide a developer guide for the Vibe-Coding block that explains how these blocks work and connect to databases, which you can also feed into an LLM.
Inside Softr, you can prompt the AI to turn specific elements into editable settings, like titles or top and bottom padding. This lets you adjust styling later via the UI without reprompting the AI. Regarding Softr's positioning between beginners and developers, the goal is enabling non-technical users to build apps via natural language, while technical profiles rely on it to deploy secure client portals quickly.
You cannot export the entire code of a Softr app because it relies completely on how Softr works internally, so standalone code would be unusable. However, having access to custom blocks gives you the freedom to inspect and understand the code. Softr's positioning is clearly for non-technical users, but offering code blocks lets advanced users push boundaries without forcing anyone to touch code.
Power users often ask about things like the Softr MCP for databases, which should arrive within the next two weeks once Anthropic connector validations are done. While we have always allowed custom code injection in headers and footers for scripts or CSS, 99% of users simply use the chat and change content titles without looking at code. You can also easily restore previous versions if needed.
You can ask the AI Co-Builder to generate an advanced component like a Gantt chart, and it will automatically add and prompt the Vibe-Coding block for you. The AI Co-Builder also works directly with data, letting you add tables or columns, and handles Softr Workflows. Users do not even need to prompt directly inside the block since the AI Co-Builder can orchestrate it across the entire app.
In our demo, we have an event brief to find speakers based on tech and design themes in the database. When tested, the webhook receives the brief and the user ID, which is the user's record ID in the database. Passing this user ID lets you log who triggered the action or associate newly created webinar records with that user.
We can configure our Softr Workflows so that after receiving the webhook brief, a step fetches all speakers from Softr Databases. Then, an AI step outputs a structured JSON listing three recommended speakers, followed by a webhook response step. While AI can draft these workflow steps automatically, building and optimizing tools like the AI Co-Builder and Vibe-Coding block behind the scenes remains an exciting challenge.
Are these AI workflows running behind the scenes? We divided the price of the Ask AI feature I showed earlier by five with simple optimizations. It is fascinating to work on building platforms like Softr and Lovable. We have the speakers table mapped, which is great. The workflow will list all speakers and pass the brief plus that speaker list to the AI so it recommends relevant speakers.
We could add a data transformation step because each speaker record has fields we might not need, like past webinars or photo size. Ideally, you clean the context passed to the AI to save tokens and improve efficiency, but we can skip that transformation for now to save time. In Softr AI, you can use direct credits or your own API key depending on your plan. We have added models like Mistral and Gemini alongside standard options, and there is now a dedicated system prompt field.
We want a structured output so that the Vibe-Coding block can receive it properly. You can either define the schema directly or provide an example. I will provide an example array here and instruct it to recommend the three most relevant speakers based on the brief in the user prompt, including their name and LinkedIn URL. Then I map the full speaker list and pass the brief received from the webhook into the prompt.
Because the structured output toggle is enabled, we get a clean JSON response. Next, we respond to the webhook with the speaker data so the AI Co-Builder or helper can handle it. I will run a test to verify everything works properly, and then I can activate the Webinar AI workflow.
This approach differs from tools like Site, which might be considered the main competitor in the vibe-coding space with its Airtable connections. In pure vibe-coding, you delegate everything entirely to the AI, whereas here you build the workflow with AI assistance and refine it yourself. That gives you full control and robustness rather than dealing with a complete black box. It might take thirty minutes instead of fifteen, but having complete visibility over your logic is incomparable to how we used to build without AI.
Pure vibe-coding tools are often constrained because they need to return a result within five minutes before a user drops off. Most people do not need overly complex setups immediately and benefit from iterating, but quick generation often skips essential details. For instance, generating an entire app from a single prompt usually misses standard 404 or 401 error pages, defaulting to generic platform placeholders whenever someone hits an invalid URL.
You do not always think about these small aspects when starting a project management tool. They are included by default in Softr so that you can focus on building business logic and structuring your database. To wrap up this part, we built a simple interface returning the top three speakers, with an AI step generating a webinar title from our prompt to create a record in the database with a description.
From here, you can build a complete experience. To show what else is possible, I have a CRM template featuring an older Softr design that I built entirely myself. We briefly mentioned themes earlier, which allow you to set an accent color applied across text, background, and surface elements throughout the app, along with simple controls for button styles, fonts, three global sizing options, and rounded corners.
Restricting design options helps maintain quality. Iker, the former lead designer at Softr, gave a talk at NoCode Week in Barcelona in 2023 explaining the challenge of offering customization while keeping tools easy to use and preventing users from making their apps look bad. Pre-suggested themes work much better than letting someone like me pick colors like bright orange.
This CRM app was missing an interface for end users to import a CSV and map columns. You can import CSV files directly in the database, but I vibe-coded an in-app interface instead. I prompted the behavior, and it performs fuzzy column matching so that fields like email and email address map automatically.
From there, you click Continue to review, select or deselect the contacts to add, and click Import Selected. This Vibe-Coding block enables completely custom features like QR code scanners, barcode readers for inventory, transcription interfaces, or AI email drafting whenever default blocks like tables and lists are not enough.
Before finishing, we should discuss Softr pricing. Rather than whether it is expensive for enterprise use, I want to address the per-user pricing model. Per-user limits were standard in no-code but feel distinct in the vibe-coding era, so what is the perspective on how user limits apply across workspaces?
By the way, if you have this plan with 500 users across 15 apps you created for your 300-employee company, it will not duplicate them. It counts unique users across the apps in your workspace. Only about 2% to 3% of workspaces exceeded 500 users, because this mainly applies to internal tools or client portals.
The real value of Softr is replacing expensive SaaS licenses. If you divide the price by the number of users, you come out ahead, replacing tools like an intranet, a deal room that costs 500 a month, HR software, or a CRM. Many users also migrate from platforms like Airtable where they paid 20 to 50 per month per user, switching to less than one euro per month per user because it fits sharing access better.
Even for a small team of three people replacing three Airtable licenses with one Softr subscription, you save money. Regarding whether you can completely replace Airtable today, using Softr Databases directly makes apps much faster and more fluid. We do not depend on external API limits, like Airtable's limit of 5 requests per second or Notion's slow API.
Airtable still has an advantage for raw data manipulation directly in the backend, as we do not have a kanban view inside the database itself since the goal is displaying it in an app. We are about to release database snapshots, which was heavily requested. We are close to feature parity with rollups, lookups, and AI fields, and this setup ensures only the right people access the right data through dedicated apps rather than raw tables.
For non-profits, NGOs, and schools using Softr for client portals, reach out to us if you need assistance. We currently offer a 30% discount and are looking into expanding that, as budgets can often be tight for those organizations.
I feel like they are starting to clutter their tool by adding AI everywhere, and it has gotten overly complex over time. At Contournement, we use Airtable heavily, but for many simple use cases with just a few tables, alternatives make more sense. Centralizing operations inside one platform was originally Glide's bet, but Softr has stayed very consistent with this approach.
In Softr, collaborators who build applications are free and unlimited. However, if you do not plan to build an app and only want a spreadsheet-style database, working directly in our database might lack certain dedicated features. I recommend Softr whenever you build apps for users, though you can also use just Softr Workflows or our unlimited free forms to replace tools like Fillout or Typeform.
The real added value of Softr remains its applications, especially internal tools and client portals. Very few tools let you manage user groups and block-level or action-level visibility so easily. It is the premier use case whenever you need to expose a slice of your data through a clean interface to internal or external stakeholders.
Thank you for this demo and the discussion around Softr's vision. After six months of major changes, the product is clearly moving in the right direction while preserving functional accessibility. It targets professional use cases, and the pricing reflects that value.
We will be at No Code Week in Frankfurt all week, and our CTO Arthur will be joining on Tuesday. Softr has always supported these community events, and we look forward to meeting people there. You can also reach out on the dedicated Softr channel within the No Code France Slack to ask questions, share use cases, or pass along product feedback.
