LiveChatAI
AIConnect AI agents to LiveChatAI, an AI chatbot platform for customer support. Agents drive conversations, post operator replies and close threads, manage the AI agents and knowledge base behind the bot, and read conversation analytics.
What This Integration Enables
LiveChatAI is worth connecting for a reason that has little to do with the chat widget. Its most interesting operation is List AI Suggestions, which returns the bot's own account of what it does not know: knowledge gaps it recorded and improvements it recommends. Paired with Add URL Data Source, Add Text Data Source, and Add Q&A Data Source, that closes a loop most support automation never closes. The bot reports where it failed, an agent assembles the answer from wherever the answer actually lives, and the knowledge base grows from evidence rather than from someone's memory of what customers ask.
The rest of the surface supports that loop. Ask AI Agent sends a question to a chatbot and returns the generated answer with conversation and message identifiers, which makes LiveChatAI usable as an answering step inside any flow rather than only on your website. Conversations can be created, listed with filters across status, sentiment, channel, resolution and assignee, and read in full. Contacts upsert by distinct ID with custom attributes, so the visitor the bot met and the customer record you already hold become one thing. Get Analytics returns conversation counts, resolution rates, response quality, contacts captured, and average chat duration over a date range with a previous-period comparison.
Two limits shape flow design and are worth stating plainly rather than discovering. There are no triggers on this connector, so nothing here fires on its own. A flow calls LiveChatAI inline, or polls List Conversations with its status, resolution and date filters on a schedule. And there is no assignment action: the connector can post an operator message with Reply to Conversation and close a thread with End Conversation, but routing a conversation to a named human happens in LiveChatAI's own inbox. A FlowRunner flow performs the handoff by notifying the right person somewhere they already work, which is usually a better outcome anyway.
Without FlowRunner
With FlowRunner
Use Case Scenarios
The bot tells you what it does not know, and a human decides the answer
A weekly flow calls List AI Suggestions for each agent in the account. Every suggestion is a place where visitors asked something the knowledge base could not cover. The flow calls List Data Sources with a search term first, because half of these gaps are wording problems rather than knowledge problems, and an existing article phrased differently is a cheaper fix than a new one. For the genuine gaps, the agent pulls the current answer out of the internal help centre or a Notion page, drafts a question and answer pair, and posts the draft to the support lead in Slack. Nothing is written until the lead answers. The approved pairs go in through Add Q&A Data Source and the rejected ones become a ticket for whoever owns the underlying product question.
An answering step inside a flow that has no chat widget in it
A field service team's dispatch flow needs to check policy questions it cannot hardcode: what the warranty covers on a discontinued part, whether a callout in a given region is chargeable. The agent calls Ask AI Agent against the internal support bot, passing the question and omitting the conversation identifier so a fresh thread is created, and requests plain text rather than Markdown so the answer drops cleanly into a dispatch note. The returned conversation identifier is stored, so a follow-up question later in the same flow can continue the thread rather than starting cold. This is the connector used as a retrieval surface over a knowledge base a support team already maintains, with no customer anywhere near it.
One customer record on both sides of the conversation
When a trial account is created in the product, the flow calls Create or Update Contact with the distinct ID already used internally, plus plan tier, signup date, and account owner as custom attributes. From then on the bot meets a known person rather than an anonymous visitor. A separate nightly flow runs the other direction: List Contacts returns the people the bot captured, List Contact Conversations pulls each one's history, and anything with a captured email that has no matching record is written into HubSpot as a new lead with the transcript attached. The support team stops copying email addresses out of chat logs.
Human-in-Loop Highlight
The gate belongs on Add Q&A Data Source, and it belongs there because of an asymmetry that is easy to miss. Every other write in this connector affects one thing: Reply to Conversation reaches one customer, End Conversation closes one thread, Create or Update Contact touches one record. A knowledge base write is different in kind. It changes what the bot says to everyone who asks that question from that moment on, in your brand voice, with nobody reading it again. There is no per-answer review downstream of it, because the entire point of the product is that the bot answers without waiting for a person. So the agent does the mechanical part, which is finding the gap through List AI Suggestions, checking List Data Sources for an existing answer, and drafting the pair from source material. Then it stops and asks: "The bot could not answer 14 conversations about refund windows on annual plans. Draft answer: [text]. Add this to the knowledge base as a Q&A pair, edit it, or reject it?" A support lead answers in Slack, and the flow calls Add Q&A Data Source only on approval. This is human-in-the-loop placed where the leverage is rather than where the risk feels highest: the agent is not asking permission to send a message, it is asking a person to own a sentence that a machine will now repeat indefinitely.
Agent Capabilities
21 actionsMessaging
3- Ask AI Agent Sends a visitor question to a LiveChatAI agent and returns the AI-generated answer along with the conversation and message identifiers. Pass an existing conversation identifier to continue a thread, or omit it to start a new one. Optionally associates the exchange with a contact and chooses whether the answer comes back as Markdown or plain text. The action that makes LiveChatAI usable as an answering step anywhere in a flow.
- Reply to Conversation Posts a human agent reply into an existing conversation on behalf of a team member, sending text, an image, or a file reference into a live thread rather than an AI-generated answer. Use it when a flow has already resolved something a person needed to decide and the customer is waiting in the thread.
- End Conversation Closes an open conversation and marks it resolved, optionally recording the human agent who ended it. Useful for closing out threads whose real resolution happened in a ticketing system.
Conversations
3- Create Conversation Creates a new empty conversation for an agent without sending an initial message. Optionally links it to a contact and tags it with an integration channel such as WhatsApp or Slack.
- List Conversations Lists an agent's conversations with filtering by status, inbox type, sentiment, channel, resolution status, assignee, and date range. This is the polling surface a flow uses in place of triggers, and the sentiment and resolution filters are what make the polling worth doing.
- Get Conversation Retrieves one conversation with its messages and metadata. Used when a flow needs the full transcript before escalating, summarising, or attaching it to a record elsewhere.
Contacts
4- Create or Update Contact Upserts a contact by distinct ID with identity fields and custom attributes. Use your own internal user identifier as the distinct ID so the bot's visitor and your customer record stay the same person.
- List Contacts Lists the contacts an agent captured. The starting point for syncing captured leads out into a CRM.
- Get Contact Retrieves one contact's profile, including the custom attributes previously written to it.
- List Contact Conversations Lists every conversation for a given contact. Used to assemble a person's full history with the bot before a human picks up the relationship.
Agents
4- List Agents Lists the AI agents, meaning chatbot instances, in the account. Each has its own knowledge base, conversations, and settings.
- Create Agent Creates a new agent. Used when a flow provisions a dedicated bot per product line, brand, or customer.
- Get Agent Retrieves one agent's configuration.
- Delete Agent Permanently deletes an agent. This takes the agent's configuration with it, so treat it as a destructive operation rather than a cleanup step.
Knowledge Base
5- List Data Sources Lists an agent's knowledge-base sources, filterable by type and search. Call it before adding anything, because a gap caused by wording is fixed differently from a gap caused by missing knowledge.
- Add URL Data Source Adds a website URL for the agent to crawl and learn from. Used to keep the bot current with a published help centre or changelog without anyone re-pasting content.
- Add Text Data Source Adds a free-form text block to the knowledge base. The right action for internal policy that exists in a document rather than on a public page.
- Add Q&A Data Source Adds a question and answer pair to the knowledge base. The most precise way to close a specific gap, and the write this page argues should never run unreviewed.
- Delete Data Sources Removes one or more data sources by ID. Used to retire content when a policy changes, which matters more than it sounds: a stale source the bot still trusts is worse than a gap.
Analytics
2- Get Analytics Returns conversation counts, resolution rates, response quality, contacts captured, and average chat duration over a date range, with a previous-period comparison. Used to give a knowledge base change a before and after rather than a hunch.
- List AI Suggestions Returns AI-generated knowledge-gap suggestions and recommended improvements. The bot's own account of where it is failing, and the natural input to a review queue.
Frequently Asked Questions
What can FlowRunner do with LiveChatAI?
FlowRunner agents can run Ask AI Agent, Reply to Conversation, and End Conversation in LiveChatAI, plus 18 more actions.
Does connecting LiveChatAI to FlowRunner require OAuth?
No. LiveChatAI connects to FlowRunner with an API key, no OAuth flow required.
Can LiveChatAI trigger a FlowRunner workflow automatically?
LiveChatAI doesn't currently expose triggers in FlowRunner. It connects as an action step inside workflows started by another trigger.
Start building with LiveChatAI
$100 in credits. No card required. Connect in minutes.