Aihio core features
Build an AI agent from your own information, publish a chat widget and improve your service using conversations.
Open copy menu
An AI agent answers customer questions using the information and instructions you provide. The chat widget is the interface visitors see on your website. Keeping these concepts separate makes both setup and troubleshooting easier.
Build, test and improve
Section titled “Build, test and improve”Give the agent a task and information
Choose one purpose, such as answering product questions. Add current sources, update the knowledge base and describe the boundaries in instructions.
Try your customers' questions
Test a familiar question and one the sources cannot answer in a preview conversation. Check the answer and any action’s result. Test messages use message credits.
Customize and publish
Set the appearance and welcome message, activate the agent and install the widget on your website. Check it on a phone too.
Improve based on what you find
Read conversations and review unanswered questions. Correct the source or instruction, retest and agree who maintains the information.
Information and response style
Section titled “Information and response style”The knowledge base supplies information for answers. You can add websites, files, questions and answers, and text. A source’s search test helps you check whether the necessary information can be found before testing a full conversation.
Instructions define the agent’s task, response style and boundaries. Templates provide an editable starting point; version history helps you return to earlier instruction text. The model and reply-language settings affect how the agent generates an answer.
Chat widget and deployment
Section titled “Chat widget and deployment”The agent’s Overview brings together its model, data sources, instructions and operating status. Appearance contains the widget’s visible settings. The preview alongside them helps you try changes.
Install a floating widget, embed a conversation within a page or use the WordPress plugin. Operating status and installation are separate: adding code to a website does not activate a draft agent.
Contact requests, human support and actions
Section titled “Contact requests, human support and actions”Lead collection offers a contact form in the situation you describe, such as a demo request. Contacts gathers captured details in a shared workspace view where you can search, filter and track follow-up status.
Human support lets a visitor request a support agent in the same conversation. When the team is unavailable, visitors can leave a message. Configure availability and agree how to handle requests before enabling it.
A custom HTTP action calls the HTTPS API you configure. For example, checking an order’s status requires a working API and its configuration; a promise written in instructions does not connect a system.
Monitoring and workspace management
Section titled “Monitoring and workspace management”| Task | What it helps you do | Guide |
|---|---|---|
| Conversation review | Find outdated information and correct recurring answer gaps. | Conversations and answer quality |
| Analytics | Review coverage and response quality to choose the next correction. | Monitor and improve the agent |
| Contact management | Pass a captured request to your team’s follow-up process. | Contacts |
| Workspace and members | Check the workspace, roles and notification settings. | Account settings |
| Usage and subscription | Monitor message credits, knowledge size and capacity. | Usage and limits |
Core features
Section titled “Core features”| Concept | What it is for | Continue here |
|---|---|---|
| Knowledge source | One web page, file, question and answer, or text source | Data sources |
| Knowledge base | The published content used to support the agent’s answers | Updating knowledge |
| Instructions | The agent’s task, boundaries and response to missing information | Agent settings |
| Chat widget | Appearance, visible text and the interface installed on your website | Publishing |
| Workspace | Your organisation’s agents, members and permissions | Account settings |
Adding a source does not train the underlying language model. The agent retrieves knowledge to support its answer. Sources and instructions still do not guarantee that every answer is correct.
Choose the next task
Section titled “Choose the next task”- Getting started: build your first agent .
- Information changed: update the source and knowledge base .
- Answers are ready: install the conversation on your website .
- Need a programmatic connection: explore custom HTTP actions .
Availability
Section titled “Availability”Available features and limits depend on the workspace subscription. Check the current options on pricing and in Oma Aihio. A connection marked as coming soon is not ready to set up.