By Lara Dashboard
AI answers from your own FAQ and docs, one-tap human handoff, a live inbox with typing previews, saved replies, internal notes and ratings.
AI support widget for LaraDashboard — answer from your products, docs, and MCPs; escalate to humans when it matters.
Drop one script tag on any page and visitors get a floating chat launcher. The AI answers first, grounded in knowledge you control. When it cannot help — or the visitor simply asks for a person — the same thread moves to a human agent with the full transcript intact. If nobody is online, the conversation becomes a lead: a CRM ticket when the CRM module is installed, or an email to your team.
Everything in this module is free. There is no paid tier, no message cap, and no locked feature.
</body>.<script
src="https://your-domain.com/livechat/widget.js"
data-site-key="lc_xxxxxxxx"
async
></script>
Replace your-domain.com with your LaraDashboard URL and lc_xxxxxxxx with the site key shown on the widget (Widgets list).
These override widget defaults for that page only:
| Attribute | Example | Purpose |
|---|---|---|
data-site-key |
lc_abc… |
Required. Identifies which widget config to load. |
data-position |
bottom-right |
bottom-right or bottom-left. |
data-primary-color |
#635bff |
Launcher and accent color. |
data-mode |
ai |
ai (assistant first) or live (human queue). |
data-open |
false |
true opens the panel on load. |
data-greeting |
Hello! |
Overrides the widget Chat welcome for this page (supports {agent}; see below). |
data-user-name |
Jane Doe |
Visitor name (hero + inbox). See Logged-in visitors. |
data-user-email |
jane@example.com |
Visitor email (inbox, offline form). |
Example with common overrides:
<script
src="https://your-domain.com/livechat/widget.js"
data-site-key="lc_xxxxxxxx"
data-position="bottom-right"
data-primary-color="#635bff"
data-mode="ai"
data-open="false"
async
></script>
The widget script is served at /livechat/widget.js (no front-end build step for embeds). The server bundles a small markdown formatter with the widget so AI and agent replies can render links and lists safely.
By default, opening the widget shows a Home hub (like a support menu), not chat immediately:
Configure under Admin → Live Chat → Widgets → Home screen:
{name} in the hero greeting for a personalized “Hi Jane 👋” when a name is known.Visitors can rate closed chats (1–5 stars) and resume threads from the Messages tab.
A conversation is only created in the inbox after the visitor sends their first message. Opening a home shortcut or the chat composer alone does not create empty threads (shortcut subject is applied when they send).
When that first message creates the thread, the widget also posts one or two welcome messages from the assistant (configured under Widgets → Copy as Chat welcome and optional Chat welcome follow-up). Use {agent} in those strings for the assistant display name (set under Admin → Live Chat → AI & Knowledge; default LaraDash). The legacy Greeting field still works as a fallback when chat welcome is empty.
Visitors can attach one file per message from the widget composer: JPEG, PNG, GIF, WebP, PDF, or TXT, up to 5 MB. Images show inline; other files link for download. Agents see the same attachments in the inbox thread.
If your site already knows who the user is, pass name and email on the script tag. The widget sends them when starting or updating a session so agents see real identity in the inbox (no extra API work on your side).
<script
src="{{ url('/livechat/widget.js') }}"
data-site-key="lc_xxxxxxxx"
data-user-name="{{ auth()->user()->full_name ?? auth()->user()->name }}"
data-user-email="{{ auth()->user()->email }}"
async
></script>
Only output these attributes when the user is authenticated; omit them for guests.
Set the widget Hero greeting to Hi {name} 👋 under Home screen so the home hub matches your reference designs.
data-user-name="Life"
data-user-email="life@example.com"
Generate those values server-side when you render the page (PHP, Node, etc.).
If identity is not available when the script loads, call this after login:
window.LivechatWidget.identify({
name: 'Life',
email: 'life@example.com',
});
This updates the home greeting and syncs name/email to the visitor record when a chat session already exists. You can also call identify before the user opens chat; the first session will include the details.
Anyone can put any email in page source, so turn on Settings → Identity verification and sign the email on your server:
$userHash = hash_hmac('sha256', $user->email, config('services.livechat.secret'));
Pass it as data-user-hash="{{ $userHash }}" or identify({ email, user_hash }). Verified visitors get a Verified badge in the inbox, and unsigned data can never overwrite a verified identity. On your own LaraDashboard site this is done for you automatically.
Privacy: Name and email in HTML attributes are visible in page source, same as other embed widgets. Only pass data you are comfortable exposing to the browser.
On the widget, Allowed origins restricts which sites may call the widget API. Leave empty to allow any origin. Set one origin per line, e.g. https://shop.example.com, when the embed lives on a different domain than LaraDashboard.
AI first, human as the safety net. The assistant answers from your connected knowledge and hands off automatically when it hits its turn limit, emits a handoff intent, or detects an escalation keyword like "refund" or "agent".
Small talk without an LLM call. Greetings, thanks, and similar short phrases are answered from a built-in catalog (config/common_phrases.php) so they never hit the “no documentation found” path or burn tokens.
Knowledge you connect. A built-in FAQ pack, plus soft integrations that light up only when the module is present: CRM products (price, SKU, description), Doc Forge documentation, and published site posts and pages. Remote MCP tools are searched with generic term matching (titles, slugs, and document names) so you are not tied to one docs product.
Multiple MCP servers. Connect any number of remote MCP endpoints — your docs export, a catalog microservice, a status page — with per-connection tool allowlists, encrypted secrets, and SSRF protection (HTTPS only, private addresses blocked unless you explicitly unlock a connection). The admin UI caches the last discovered tool list per connection. A failing server is skipped, never fatal.
Realtime you choose. Polling is the default and works on shared hosting with no daemon. Switch to Laravel Reverb or Pusher in settings; Live Chat falls back to polling automatically if the selected driver is not fully configured.
Deep customization. Colors, launcher size and position, corner radius, light/dark/system theme, custom CSS, greeting and every other string, GDPR consent gate, identity capture rules, and configurable home shortcuts — all per widget. Powered by LaraDashboard footer branding is off by default; turn it on under Admin → Live Chat → Settings if you want it.
One AI integration for the whole app. Live Chat uses LaraDashboard Settings → Integrations for API keys and models. The Live Chat AI & Knowledge screen only toggles whether the assistant runs and how it behaves (system prompt, fallbacks, turn limits, assistant display name) — it does not store duplicate provider credentials.
Readable AI and agent replies. Assistant and human messages support lightweight markdown (links, lists, code spans) in both the widget and the admin thread.
Agent inbox. Filter by open, waiting-for-human, assigned to you, unassigned, or closed. Reply with text or an attachment, claim, release, close, reopen, and bulk-close selected threads. Saved Replies support optional shortcuts (e.g. /refund): type the shortcut and press Space in the composer to expand, or send the shortcut alone. Presence heartbeats tell the widget whether anyone is actually around.
Internal notes. Switch the composer to Internal note to leave context for teammates. Notes are amber in the thread and are never sent to the visitor or the AI.
Typing indicators, both ways. Visitors see “Sam is typing…”, and agents see a live preview of what the visitor is writing so they can start answering early.
Business hours. Set opening hours per weekday and timezone. Outside them the AI keeps answering, the widget says when the team is back, and “Talk to a human” collects an email instead of queueing.
Tidy inbox, safe data. Idle chats close automatically after the hours you choose, and a retention window deletes old conversations together with their uploaded files (livechat:maintain, hourly via the scheduler).
Built for the public internet. Per-visitor daily AI reply budget and upload limits stop one visitor from burning credits or filling your disk. Every widget endpoint honours the origin allow-list, the visitor token travels in a header (not the URL), and realtime events are signal-only on a private agent channel.
Visitor context. The thread header shows the page the visitor is on and whether their identity is verified.
Operator MCP tools. Run support from your own AI agent with list-livechat-conversations, get-livechat-conversation, reply-livechat, close-livechat, and list-livechat-mcp-connections. Queued and unread counts also show up in the daily briefing.
The widget ships with keyboard send, Escape to close, aria-expanded on the launcher, a live region for incoming messages, visible focus rings, and a full-width mobile sheet.
livechat.inbox, livechat.reply, livechat.assign, livechat.settings, livechat.canned, livechat.ai, livechat.mcp
An agent without livechat.settings or livechat.ai never sees your API keys or MCP secrets.
No reviews yet.
Purchase this module to be the first to review!
Released Nov 17, 2026
No changelog available for this version.
0
Downloads
0
Active Installs
0.0
Average Rating
0
Reviews
/shortcuts, and "Add to FAQ" from any reply.
livechat:maintain, hourly).
Lara Dashboard
Module Developer