AI chatbot for your website with AI Flowi Workflow

By Ai Flowi · Published · Updated

You can offer an AI chatbot for your website by building a chat workflow in AI Flowi Workflow, publishing it at its own web address and linking visitors to it. Plans: Go from USD8/month billed yearly (USD10 monthly), five plans up to Pro 500. Every plan is listed on the pricing page. The publishing guide was published 7 October 2026; pricing checked 1 November 2026.

AI Flowi Workflow is a visual workflow builder from Ai Flowi (aiflowi.com, operated by AI FLOWI SDN. BHD.) for building AI agents, chatbots and automations from drag-and-drop components. Plans: Go from USD8/month billed yearly (USD10 monthly), five plans up to Pro 500. Every plan is listed on the pricing page. Each node run uses 1 Flowi Token.

How it works

Start with the documented chat path: connect Chat Input to a compatible AI model, then connect the model to Chat Output. Configure the model’s required credential and send a message in the Playground to check the response. Matching component types establishes a compatible connection; the response still depends on the model configuration and service.

If the chatbot needs information from your documents, add a knowledge base and retrieve relevant content inside the workflow. Review the retrieved content before using it in an answer. Document retrieval is optional: the first-workflow tutorial starts with Chat Input, a model and Chat Output.

Publishing links the saved workflow to a visitor-facing web app. Visitors open the published address, enter text and receive the workflow’s text reply. Use the supplied app address or configure a custom domain, then add a descriptive link from your website to that address.

The publishing documentation describes a separate web app. No embeddable website widget or script snippet is documented in these guides. Plan your website link around the published app address.

Steps to publish your chat app

  1. Choose the workflow and app type. Save and test your workflow. Open its Publish Web action, or start a new app in the published-app list, and select Chat. The wizard also lists form, dashboard, portal and custom app types. Once an app draft exists, its workflow and app type are locked.
  2. Set the address and access. Choose the supplied address or a custom domain, enter the address name and page title, and select public, login-required or private-link access. For a custom domain, its administrator must configure the supplied direct CNAME target and TXT verification record, complete DNS verification and ensure TLS is deployed.
  3. Add branding and review. Enter the required brand name and heading. You can also set an introduction, colours and suggested prompts, and add a logo and favicon. Check the preview and fix validation errors before publishing.
  4. Publish, test and link. Review the saved settings and select Publish Web. Open the address as a visitor and send a harmless test message. Check the reply and the chosen access restrictions before adding the link to your website.

Choose access to match who should use the app:

Access modeVisitor requirement
PublicAccepts visitor runs without a workspace login.
Login-requiredChecks the signed-in user’s workspace permission.
Private-linkChecks the issued link token. Private links can be rotated or revoked.

Before publishing, check your workspace’s publishing access and published-app quota. After editing settings, publish again to replace the published version. Follow the publishing guide for the complete checks.

Answering from your own documents

Knowledge bases store document content for search and retrieval inside a workflow. Open Knowledge, select Add Knowledge, name the base and choose an embedding model. Configure Sources lets you upload files or a folder. Accepted formats include PDF, DOCX, TXT, Markdown and CSV; unsupported files are skipped and listed in an alert.

Review Chunk Size, Chunk Overlap and Separator, then inspect Chunk Preview before creating the base. Wait for Ready in the knowledge list: closing the upload dialog does not mean ingestion has finished. Use View Chunks to inspect the stored text.

Add the Knowledge Base component to your workflow, select the base and enter a Search Query. Set Top K Results and Include Metadata as needed. Run the component and inspect its Results table. Test with a question covered by your documents and compare the retrieved content with View Chunks before testing the chatbot’s answer.

The knowledge-base guide documents setup and retrieval. It does not establish a completed knowledge-base-to-model-to-published-chat test. Validate that full path with your own workflow before sharing it.

Cost

Each node run uses 1 Flowi Token. Plans: Go from USD8/month billed yearly (USD10 monthly), five plans up to Pro 500. Every plan is listed on the pricing page.

Third-party API usage is not included in the AI Flowi subscription.

The chat workflow uses an AI model, so review the selected model provider’s charges alongside the subscription. A knowledge base also requires an embedding model and any credentials that provider needs. Check AI Flowi Workflow pricing and test your configured workflow before estimating its usage.

Sources: First workflow tutorial, publishing guide, knowledge-base guide, publishing features and pricing.