Learn Flexaura, one step at a time.

Steps

1. Sign up at app.flexaura.ai. You land on the onboarding page.
2. Name your workspace. Your agents, knowledge bases and conversations all belong to it.
3. Open Settings → Organization to check the workspace details.
4. Open Settings → Team members and invite colleagues by email.
5. If you belong to more than one workspace, switch between them at the top of the sidebar.

Steps

1. Open Templates in the sidebar. Under Generative agents, read each description and click Show instructions to see what it tells the AI.
2. Click Use template on the closest fit, for example Customer Support or FAQ Bot.
3. Basics: give the agent a name and check the suggested display name visitors will see.
4. Personality: read the instructions and adjust the three tone sliders.
5. Model: pick an AI model. If you're unsure, keep the default.
6. Continue through Knowledge and Channels (tutorials B3 and B5), then Publish.

Steps

1. Open Knowledge Bases and create one, for example "Help centre".
2. Upload documents: .pdf, .docx, .txt or .md, up to 25 MB each.
3. To use your website, click Add from a URL instead, then add one page with Add URL or a whole site with Crawl site.
4. Wait until every document has finished processing.
5. Edit your agent, choose this knowledge base on the Knowledge step, and save.

Steps

1. On the Agents page, open the agent's menu and choose Playground.
2. Ask five real questions your customers ask, plus one the agent shouldn't be able to answer.
3. Where an answer is off, tighten the instructions on the Personality step, or add a document to the knowledge base.
4. Clear the chat to start a fresh conversation after each change.

Steps

1. On the Channels step, keep Web chat on and set the colour, style, corner and welcome message.
2. On the Publish step, open the JavaScript tab and copy the snippet.
3. Paste it just before </body> on every page that should show the chat, or into your site builder's custom code setting.
4. Publish your site and open it in a private browser window.

Steps

1. Open Conversations and pick one.
2. Click Take over. The agent stops replying in that conversation.
3. Type your reply. The visitor sees it labelled "Support agent".
4. When you're done, click Return to bot, or mark the conversation resolved.

Steps

1. Open Analytics for the whole workspace, or choose Analytics from one agent's menu.
2. Change the date range to compare weeks.
3. Look at conversations per agent. For flow agents, also see where visitors stop and why they were handed to a person.
4. Turn what you find into a fix: a missing document, or a clearer instruction.

Steps

1. Check the plan card on the Dashboard for this month's usage.
2. Open Settings → Billing & plan to compare Free, Pro, Team and Enterprise.
3. Upgrade through checkout. New features are available straight away.

Steps

1. Edit the agent and go to the Personality step.
2. Say who the agent works for, what it must never do (for example, promise refunds), and when it should offer a person.
3. List the topics it should politely decline.
4. Save, then ask your Playground questions from B4 again.

Steps

1. On the agent's Publish step, open the React tab for React apps: install @flexaura/react and render the FlexauraWidget component exactly as the tab shows.
2. Open the iFrame tab for a chat that fills a space on the page, such as a help page.
3. Try either one on a test page before your live site.

Steps

1. Open Email Configs and create one. You get a Flexaura email address for it.
2. Attach an agent. Mail sent to an address with no agent attached is ignored.
3. In your own mailbox settings, forward your support address to the Flexaura address.
4. Optionally, add your own domain so replies come from your address. This needs DNS records at your domain provider.

Steps

1. Open the agent's menu and choose Workflow.
2. Switch on what you need: Search the knowledge base, Remember details, Offer reply buttons, Hand off to a human, Resolve conversations.
3. Save, then try each one in the Playground. For example, ask for a person and check the conversation moves to your team.

Steps

1. Create an API key in your AI provider's console.
2. Open Settings → LLM keys (BYOK) and add it. It is stored encrypted, and only its last few characters are shown again.
3. On the agent's Model step, choose one of that provider's models.

Steps

1. Know the default: every incoming message is checked for harmful content before the agent replies, and a flagged message gets no AI reply.
2. In the instructions, say when to offer a person: an angry customer, a billing dispute, anything legal.
3. On Pro and above, switch on Hand off to a human (I4) so the agent can move the conversation itself.
4. Decide who on your team watches Conversations for handovers, and when.

Steps

1. In Settings → Team members, choose Admin or Member for each person you invite.
2. Revoke invites you no longer need.
3. For help with setup, invite a Flexaura support address into your workspace. It doesn't use a seat, and you can remove it at any time.

Steps

1. Open Templates and, under Flow agents, pick one such as Callback request or Order status.
2. Open the card's "things to fill in" list to see what you'll need before it can go live.
3. Click Use template, choose a name and how the chat looks, and create the agent. The flow editor opens.
4. Work through the issues list: replace the example email addresses and api.example.com addresses, and add any secrets it names.
5. Run the template's saved tests, then publish.

Steps

1. Open Settings → Tools and start from a template (Shopify, WooCommerce, HubSpot, Zendesk, Calendly bookings, Send a lead) or from scratch.
2. Set the HTTPS address, the method, and the inputs the AI should fill in, with a clear description of each.
3. Add your API key as a secret. It is encrypted and never shown in full again.
4. For anything that changes data, switch on Ask the visitor before running: the agent describes the action, and it runs only after the visitor replies.
5. Attach the tool on the agent's Workflow page and try it in the Playground.

Steps

1. Sketch the process first: what you ask, what you check, and where it ends.
2. Create a Deterministic agent and choose Blank as the starting point.
3. Add steps: Message, Buttons, Ask for (collects a value), Confirm, Condition or Switch, API call, Hand off and End.
4. Move parts you repeat into a Sub-flow, and use Go to to return to a menu.
5. Add keyword commands such as "menu" or "agent" that work at any point.
6. Add Mark goal where the flow succeeds, so analytics can count it.

Steps

1. Run a conversation in the test panel and click Keep this conversation as a saved test.
2. Build up saved tests until each main path has one.
3. Edit the draft, run all saved tests, and publish only when they pass.
4. Before publishing, read the list of changes since the live version.
5. If something goes wrong, open the version history and restore the previous version.

Steps

1. Choose Hybrid when you create the agent, and set it up like an AI agent: instructions, model and knowledge base.
2. On its flow page, add the tasks that must follow an exact path, such as returns, each started by an intent.
3. Anything that matches no intent is answered by the AI.
4. Add an AI answer step inside a flow when you want an answer from the knowledge base at that point.

Steps

1. For each intent, add keywords and several example phrases in your customers' own words.
2. Review the settings for typos, "Did you mean…", switching between tasks, and synonyms.
3. Turn on matching by meaning to catch phrasing your examples don't cover.
4. After launch, check "How messages were understood" in the agent's analytics and add missed phrases to the right intent.

Before you start

You need a developer who can change your website's server code.

Steps

1. On the agent's flow page, open Verified visitors and create a secret. Store it on your server only.
2. For each logged-in customer, your server signs a JWT with that secret using HS256. It must contain sub (your customer's ID) and exp, at most one hour ahead. name, email and attrs are optional.
3. Give the token to the chat: set userToken (or getUserToken, to fetch a fresh one when it expires) in FlexauraConfig for the JavaScript snippet, or as props on FlexauraWidget. Call Flexaura.identify(token) after a customer logs in on the page.
4. Put a Verified only step before anything account-specific, such as freezing a card.
5. Optionally switch on Only verified visitors can chat.

Steps

1. Open a Deterministic agent's flow page and find the AI usage panel.
2. Change the safety check on incoming messages from OpenAI to keywords (checked on Flexaura's own servers) or off.
3. Read the statement at the top of the panel. It lists exactly which providers see this agent's messages.
4. Remove any AI answer steps if no AI provider may be involved.

Steps

1. In the Playground or the conversation, open the reply's tool activity: which tool ran, with what inputs, and what came back.
2. Check the reply's timings to see whether searching, the AI model or a tool call was slow.
3. Check this month's tool-call usage, on the agent's Workflow page or in Analytics, against your plan's limit.
4. For a flow, use "Show the steps it took on the canvas" to replay the path a conversation took.