What changed. Why it matters.

The home-page product demo now shows more clearly how company knowledge is prepared, how the agent matches the company’s brand, and how an answer and its sources are reviewed before publishing.

• Improvement
  The Train step shows switchable Websites, Files, and Q&A views for the future Knowledge Base. Website processing moves through a four-step tracker described in buyer-friendly language rather than implementation terms.
• Improvement
  The Customise step now mirrors the real editor with tabs and expandable setting groups, while showing self-service editing for the name, logo, theme, primary colour, corner shape, window size, and placement. The widget preview is now larger.
• Improvement
  The Test step continues with the same agent and prepared knowledge base. A sample question progresses through knowledge retrieval to an answer whose source, supporting excerpt, and all three sources can be reviewed at leisure.
• Improvement
  The lifecycle presentation can be paused, and the Customise step remains visible for longer. Choosing a setting pauses the presentation so the result can be reviewed at leisure.

Site menus now describe their destinations more clearly, while cookie preferences present consent choices with equal prominence.

• Improvement
  Header menus now include a short description for each destination on desktop and mobile. The menus close with Escape and return focus to their trigger.
• Improvement
  The cookie banner presents acceptance and rejection as equal choices. Visitors can review detailed preferences before deciding and change them later.

Chatbot owners can limit the widget interface languages and let visitors switch between Finnish and English.

• New
  Owners can choose the interface languages allowed in the widget and enable a visitor language picker in the Appearance section of the editor. Free
• Improvement
  A visitor’s language choice persists across visits while always respecting the owner’s current language allowlist.
• Improvement
  The new Settings → Languages view brings reply language, widget interface language, and knowledge search language together. Read more →

Card-rendering actions, such as the contact form, now run only on channels that can display a card. Operators can restrict what an action exposes and require visitor confirmation before it runs.

• Fix
  Card-rendering actions are now registered only on channels that can display a card – on the REST API and text-message channels the bot keeps answering in plain text instead of leaving the conversation waiting on an invisible card.
• Improvement
  Operators can restrict what the AI sees from an action’s API response. Starter
• Improvement
  An action can require visitor confirmation before it runs, and every run is recorded in a run log.
• Improvement
  Configuration changes to an action are now versioned in its history.
• Fix
  If a human-support request goes unanswered, the AI resumes the conversation after thirty minutes of silence – the conversation no longer waits indefinitely.

Code-styled text and form validation messages now retain their colour contrast. The chatbot switcher has a larger touch target on narrow screens.

• Fix
  Code typography no longer removes the text colour, improving the readability of code blocks and form validation messages.
• Fix
  The chatbot switcher now meets the WCAG 2.2 target-size requirement on narrow screens.

Errors, suggestions and sources now sit with the message they belong to instead of taking a row of their own. A set of accessibility gaps closed with them, along with a bug that kept the chatbot’s own uncertainty note from ever appearing.

• Improvement
  A failed message reports the error in its own place, not as a separate bar above the composer. The bubble is highlighted only when the message never reached the server — if the answer broke mid-reply, it did arrive and is not worth sending again
• Improvement
  Running out of credits, or sending too quickly, is now explained to the visitor. Before they saw only a generic error with no way forward
• Improvement
  Follow-up suggestions appear on the visitor’s own side and look like the message they will send. The same sentence used to appear twice, on two different sides
• Improvement
  Sources fit on the message’s own row beside the timestamp. The list still expands on click, but no longer costs a separate row under every sourced answer
• Fix
  The source list could come up completely empty. When one document produced several matches in an answer, those sources shared an identifier and the list was never rendered at all — while the control still reported itself as open. Every source now shows
• Fix
  The chatbot’s note about an uncertain answer never appeared at all under default settings. It now sits behind a small icon — an owner-written note is still always shown in full, because it may be a required legal notice
• Fix
  An interrupted answer showed the date 1.1.1970 underneath it
• Improvement
  Accessibility: the transcript can be scrolled with the keyboard, the reaction and copy controls meet WCAG 2.2 target sizes, the launcher’s tooltip is reachable by keyboard, and the sensitive-data hint no longer interrupts a screen reader

Verification and notification emails now use the user’s selected language. Mobile onboarding gives clearer guidance for adding documents, testing the chatbot, and choosing its appearance.

• Fix
  Email subjects no longer expose unresolved variables, and messages are sent in both HTML and plain text in the user’s language. Free
• Improvement
  Lead and conversation-request emails follow each member’s notification settings and verify membership before delivery. Starter
• Improvement
  Onboarding can combine a website, local files, and question-and-answer pairs. The chatbot preview and installation actions now fit mobile screens more clearly. Free
• Improvement
  The chat widget launcher can use a colour separate from the primary colour. Icon contrast is selected automatically. Starter

Channels shows where customers can talk with the chatbot. Chat-widget installation opens directly from the Editor, while technical security settings have a dedicated view.

• Improvement
  The Editor’s Install menu provides the chat widget, inline chat embed, and WordPress plugin guidance. Free Read more →
• Improvement
  Channels brings together the chat widget, shareable chat page, and text messages. Integrations is reserved for external systems, automation, and APIs. Free
• Improvement
  Domain restrictions and authenticated-user identity settings are under Settings → Security. Free Read more →

When the same agent sends several messages in a row, they now read as one group, the same way the bot’s messages already did. Each message used to repeat the name and the time, which broke up the conversation.

• Improvement
  Consecutive agent replies are grouped: the avatar appears only on the first message, the name and timestamp only on the last, and every following bubble keeps a squared top corner

You can now type freely in the command palette (Cmd/Ctrl+K). The letters h, i, v, s and a used to jump straight to another page mid-word, so a search could never start with them.

• Fix
  Command palette search accepts every letter. h, i, v, s and a used to navigate away immediately, which made it impossible to search for terms starting with them
• Improvement
  Those letters still work as shortcuts when the command palette is closed and the caret is not in a text field

Solutions, Enterprise, pricing, integrations, and resource pages are now structured around their own jobs. Product views demonstrate the knowledge base, current-page context, handoff, and lead capture.

• Improvement
  The new Solutions menu brings use cases, features, integrations, and enterprise services together. Read more →
• Improvement
  Pricing, integrations, and frequently asked questions now lead more directly to a plan decision, an available connection, or the right answer. Read more →
• Improvement
  Documentation has a new starting hub, while the blog and changelog use a clearer editorial structure and the real RSS feed. Read more →

The redesigned home page shows how to build, test, publish, and improve an AI agent grounded in your own content by yourself. Its structure, typography, and product views now work consistently from mobile to wide screens.

• Improvement
  You can try the chat widget’s colour, theme, launcher, and position in a live preview, then explore the agent’s knowledge, conversations, use cases, data controls, and pricing on the same page. Read more →

The conversation list, message history, and customer details remain usable in long conversations. Channel filtering and forms shown in the widget now render correctly.

• Fix
  The workspace Conversations link opens the shared inbox again. The list, message history, and customer details now scroll independently. Pro
• Fix
  The channel filter shows a clear All channels selection and only channels used or active in the workspace. Pro
• Fix
  Lead forms shown in the widget appear as events in conversation history without exposing submitted form values. Pro
• Fix
  The widget no longer shows anonymous visitors an unnecessary notice about device-local conversation storage. Free

Aihio returns users to their last active workspace and avoids unnecessary onboarding in an empty personal workspace when a suitable team workspace is already in use.

• Fix
  Workspaces can also be switched during onboarding. The application root now opens sign-in or the user’s workspace instead of the marketing site. Free

Dashboard views, chatbot settings, and setup steps now guide users more consistently to the next task. Save status, recovery guidance, and technical options appear in the right context.

• Improvement
  Publishing, Installation, and Security are now separate tasks. Readiness, preview, installation guidance, and security next steps are available in one coherent flow. Free Read more →
• Improvement
  Chatbot settings explain more clearly what has changed and when changes were saved. Technical options remain available under advanced settings. Free Read more →
• Improvement
  Knowledge, integrations, analytics, conversations, contacts, team, and billing views now use more consistent actions, states, and recovery guidance. Free
• Improvement
  Mobile navigation keeps the active chatbot context without duplicate menus, while loading and error states follow the structure of their destination views. Free

The editor protects drafts from concurrent changes, while the embedded widget recovers from failures more clearly and stays isolated from host-page styles.

• Improvement
  The editor saves only changed settings. If the same chatbot is edited in another tab, you can resolve overlapping changes without losing your draft. Free Read more →
• Fix
  One invalid legacy setting no longer resets the rest of the widget configuration, and visitors can retry a loading failure directly in the widget. Free
• Fix
  Widget typography and text styles no longer inherit CSS from the host page. Colors and themes selected in the editor still remain intact. Free
• Improvement
  Builder navigation uses less space on mobile, and human-support presence appears only when human support is enabled. Free

Optional page recognition helps the chatbot answer referential questions such as “What does this include?” using the web page where the visitor sends the message.

• New
  Current page recognition is available as a beta feature on every plan. Enable it in the chatbot AI settings without changing the embed code. Free Read more →
• Improvement
  The chat widget sends only an identifier derived from the page URL in the browser. It does not send the raw URL, query parameters, page text, or form data. An unrecognised page uses normal knowledge search. Free

Every workspace page now shares one light header and a centred fixed-width content area, so Overview, Billing, Team, Settings, and Contacts all feel like the same product. The Overview page is now the hub: the chatbot list and usage analytics are folded in, so the separate Chatbots and Usage pages are gone. Geist font throughout the dashboard.

• Improvement
  All workspace pages share the same light page header and fixed-width content area; key figures use consistent KPI stat cards; in-page tabs (Billing, Settings) now use the clean underline style
• Improvement
  Overview is now the hub: chatbot list as large clickable cards at the top, and usage analytics filterable by period (7/30/90 days) below; the standalone Chatbots and Usage pages are removed and their old URLs redirect to the hub automatically
• Improvement
  A header "Open chatbot" quick-picker gives direct access to any chatbot editor without navigating away
• Improvement
  The command palette (Cmd/Ctrl+K) is modernised: match highlighting as you type, chatbot avatars, timestamps on recent items, and keyboard hints at the bottom

Invites, one-time passcodes, and notifications now follow the current brand, support dark mode, and are more accessible. We also fixed an issue that stopped team-invite emails from sending.

• Fix
  Team-invite emails send again; they could previously fail silently
• Improvement
  Transactional emails refreshed to the current purple brand, with dark-mode support and accessibility improvements (WCAG 2.2 AA)
• Improvement
  Emails now include a plain-text part for better deliverability and accessibility

When your chatbot captures a new lead, you're notified right away: by email, in the dashboard, and optionally to your own system through a signed webhook. Turn notifications on or off per chatbot. Leads no longer sit unnoticed in your contacts list. Available under a chatbot's Settings → Leads (Starter tier and up).

• New
  Email notification of a new lead to the account owner, with a branded template
• New
  In-app dashboard notification with a direct link to the captured contact
• New
  Signed lead.captured webhook to your own system or Zapier, with the lead details included
• Improvement
  Per-chatbot toggle for lead notifications, on by default

Define an HTTPS API call that the AI agent fills in and the server runs during a conversation, such as looking up an order, booking a time, or creating a ticket in your own system. Credentials are stored encrypted, calls run server-side with SSRF protection, and you can optionally pass a verified visitor's details along. Available under a chatbot's Settings → Actions (Starter tier and up).

• New
  Operator-defined HTTPS action: the AI agent fills in the parameters and the server runs the call during the conversation
• New
  Credentials (Bearer or API key) stored encrypted; calls are https-only and SSRF-protected
• New
  Optional pass-through of a verified visitor's details, with a GDPR warning
• New
  The chat widget shows visitors when the AI agent is performing an action

You can now cryptographically verify a logged-in user's identity in the chat widget: a per-chatbot secret generates HMAC-SHA256 user hashes or signed HS256 JWT tokens, so the AI agent knows exactly who it is talking to before handling personal data. Compatible with common identity-verification models. Key management is available in the chatbot's Security view.

• New
  HMAC-SHA256 and JWT-based identity verification for the chat widget
• New
  Key management in the Deploy view: create, rotate, and delete the verification secret per chatbot
• Improvement
  Signed tokens can carry user attributes (e.g. customer identifiers) for upcoming identity-scoped actions

The AI agent now answers SMS too. Connect your own Twilio account in a few clicks — you own the numbers and pay Twilio directly, Aihio never relays the bill.

• New
  BYO Twilio SMS channel: Account SID + Auth Token + phone number from the customer's own Twilio account
• New
  Inbound SMS is routed to the same AI agent that answers the website chat widget
• New
  Channel badge + channel filter in the conversation inbox
• New
  Per-reply delivery status (sent / failed + reason) so operators see what reached the visitor
• Improvement
  Credentials stored AES-256-GCM encrypted at rest; Twilio webhook signatures verified on every inbound message

A real-time accessibility checker makes sure the chat widget's colors meet WCAG 2.2 AA requirements. One click fixes the contrast.

• New
  Real-time WCAG 2.2 AA contrast warning in the color picker
• New
  Automatic fix button picks the nearest compliant shade
• New
  AA-Large distinction for large text elements (18 pt / 14 pt bold)
• Improvement
  EAA readiness for Finnish AI agents

Connect your AI agent to your own systems: send messages through the API and receive events via webhooks. Available from the Starter plan.

• New
  REST API for sending and receiving messages
• New
  Webhook events: new_message, conversation_ended, and escalation
• New
  API key management in the dashboard (create, rotate, delete)
• Improvement
  HMAC signature on webhook requests for security

Voice input, multi-message replies, and grouped message rendering make the chat widget clearer.

• New
  Voice input via microphone (Groq Whisper, Finnish supported)
• New
  Multi-message replies: the agent can split a long answer into several bubbles
• Improvement
  Grouped message rendering with a clearer visual hierarchy
• Improvement
  Updated AI labeling for better transparency

Welcome to Aihio's first public beta. The self-service AI agent builder is now available.

• New
  Build an agent from a website URL or uploaded documents
• New
  Finnish-language AI chat, with multilingual support for the Nordic languages
• New
  Dashboard for managing the knowledge base, settings, and analytics
• New
  Data processing built in the EU