Frequently asked questions

About 5 minutes for a working agent: sign in with Google, create an agent, point it at your website URL (or upload documents), and paste one small script tag into your site. Tuning the tone, lead-capture questions, and appearance is optional and can be done any time after launch.

No. Installation is copy-pasting a single script tag — if you can edit your website (or your website builder has a "custom code" box), you can install Dialogent. If you'd rather not touch your site at all, every agent can also get a hosted chat page on dialogent.io that you simply link to.

Any platform that lets you add a script tag or custom-code embed: WordPress, Squarespace, Wix, Shopify, Webflow, Framer, and hand-built sites all work. The widget is a standard embed — there's no platform-specific plugin to install.

Yes. Every agent can be published as a hosted, branded chat page at dialogent.io/a/your-business — share that link from your Instagram bio, Google Business Profile, or email signature and it works exactly like the website widget.

Yes — the live demo has example agents for a dental clinic, restaurant, law firm, fitness studio, real-estate brokerage and more. It's the real product, not a video — the floating appearance even embeds the exact widget you'd install on your own site, lead-capture form included. The Free plan also requires no credit card, so you can build your own agent and test it end-to-end at no cost.

Yes. You can chat with your agent privately in the dashboard before embedding it, and the Knowledge tab has a search tool that shows exactly what the agent knows about any question — so you can verify answers about your prices, hours, and services before going live.

Yes. Each agent has its own knowledge, playbook, and widget, so the usual setup is one agent per website or location. The Free plan includes 1 agent, Starter 3, Pro 10, and Enterprise is unlimited.

Three source types: website URLs (we fetch and index the page), pasted text (great for FAQs, policies, or service menus), and file uploads — PDF, DOCX, TXT, Markdown, and HTML. Most businesses start with their website plus a price list or FAQ document.

Dialogent is built to answer only from your content. Every reply is grounded in retrieved passages from your knowledge sources and shows citations, and an output guardrail checks answers for grounding before they're sent. If your content doesn't cover a question, the agent says so instead of guessing.

It tells the visitor it doesn't have that information and offers to connect them with your team (capturing their contact details). Every unanswered question is also logged as a "knowledge gap" in your analytics, so you can see what to add to your content next.

Update the source and re-sync it with one click — pasted text can be edited directly, and URL sources re-fetch your live page. Changes are indexed within minutes. Automatic scheduled re-crawling isn't built yet, so plan to re-sync URL sources after you change your site.

The agent runs on OpenAI models, which understand most major languages, and it will generally reply in the language the visitor writes in — if your knowledge content covers the answer. Dialogent's own dashboard and per-language controls are English-first today.

Yes. The agent's playbook lets you set explicit rules ("never give medical advice", "always include this disclaimer"), and built-in guardrails check every incoming message for relevance and safety. Off-limits questions get a polite redirect or a human handoff instead of an answer.

Yes. Replies include a sources footer citing the pages or documents the answer was drawn from — visitors can verify, and you can audit exactly which source produced any answer.

Yes — paste them as a text source. Explicit Q&A pairs are the most reliable way to control the exact answer to a high-stakes question (refund policy, insurance accepted, service area).

Two ways, and you choose the style per agent. With the inline style, the widget shows a small form (name, email, phone — you pick which fields) after the first reply; with the conversational style, the agent asks for contact details naturally in the conversation instead (no form). Either way, Dialogent also automatically extracts lead details the visitor mentions — contact info, company, intent, budget, timeline — so you get a qualified lead profile even when nobody fills in a form.

With the built-in Square Bookings integration, the agent books for real: it lists your services with their prices, checks live availability, and creates the booking right in the conversation — including at 11pm when nobody's at the desk. It handles add-on services in the same appointment, and if your business travels to the customer (home services, mobile detailing), it collects the service address and puts it on the booking. The visitor gets an email confirmation with the service, time (in your business's timezone), and team member (when they've left an email address), and workspace admins get a booking alert.

Yes — with the built-in Booking requests integration (no scheduling system needed). The agent collects the service, the visitor's preferred times, and their contact details in the chat, then emails the request to your team instantly; you confirm the final time with the visitor directly. The visitor gets a "request sent" confirmation (and an email receipt when they've left an address). Enable it per agent in the dashboard — there's nothing to connect.

Yes. You connect your own Square account once in the dashboard — sign in with Square (OAuth) or paste an access token — pick the location, and enable booking per agent. The agent reads your real services, variations, and prices from Square and quotes them in chat — no re-entering your service menu. Mobile and at-customer-location setups are supported: the agent collects the visitor's address and it lands on the Square booking.

Yes — built in. Connect HubSpot once in the dashboard with a private-app access token (scopes: crm.objects.contacts.read and crm.objects.contacts.write), enable it per agent, and every captured lead is synced to HubSpot as a contact the moment it's captured — name, email, phone, and, for conversation-extracted leads, company and job title too. Contacts are upserted by email address, so repeat visitors update their existing contact instead of creating duplicates. Works for both form-captured and conversation-captured leads.

Yes — with the built-in Google Calendar integration, every appointment booked in chat is mirrored onto your calendar automatically: an event is created on booking, moved when the visitor reschedules, and removed when they cancel. Connecting takes two steps and no technical setup: share your calendar with Dialogent's calendar address ("Make changes to events" — the dashboard shows you the address to add) and paste your Calendar ID. The event includes the service, customer name and email, and team member; to disconnect, just unshare the calendar. Enable it per agent alongside Square Bookings.

The built-in integrations today are Square Bookings (scheduling), Booking requests (email — works with any calendar), HubSpot (lead sync), and Google Calendar (booking sync). Calendly-style flows, Outlook calendars, and other CRMs aren't built in yet — for those, signed outbound webhooks fire on every captured lead and every booking made or requested in chat, so you (or a tool like Zapier/Make) can push them into any system. There's also a "request an integration" form in the dashboard — requests directly shape what we build next.

Workspace admins get an email alert the moment a lead is captured, the lead syncs straight into HubSpot as a contact when you've connected it, webhooks can notify any system you run in real time, and every conversation with its lead profile is in the Sessions dashboard.

Yes — that's the core of the product. The agent answers, qualifies, and books 24/7. Businesses replacing after-hours answering services or missed calls are exactly who Dialogent is built for.

Yes. The playbook defines what the agent should learn from a visitor (case type, urgency, budget, service area — whatever qualifies a lead for your business) and when to ask for contact details, and the conversation extraction fills in whatever the visitor volunteers along the way.

Yes. The visitor gives the email (or phone) they booked with, the agent looks up their upcoming appointments, and can move one to a new open slot or cancel it — always after an explicit confirmation. The contact details double as the ownership check, so nobody can touch a booking that isn't theirs, and both sides get an email notice of the change.

Yes — colors, launcher icon, agent avatar, corner radius, button labels, and the greeting are all configurable per agent, so the chat looks like part of your site rather than a bolted-on tool. The widget tab can also read your website and suggest matching colors automatically, and every widget supports light and dark palettes — set it to auto and it follows each visitor's system theme.

Yes. The widget can wait to be clicked, auto-open after a delay you set, open on exit intent, or stay hidden on mobile entirely. Sessions only start when a visitor actually opens the chat, so passive page views never count against your plan.

Yes. The widget resumes the same conversation across reloads and page navigations on your site — the transcript comes back and the agent keeps its context. A conversation stays resumable for 60 minutes after the visitor's last message, then the next open starts fresh; visitors can also tap the restart button in the chat header to start over at any time. Resuming never counts as a new conversation against your plan.

Yes. Use "Pause agent" (in the agent's menu, on the agent page or the agents list). Pausing stops the agent everywhere at once — the website widget stops appearing, and the hosted chat page shows an "unavailable" notice — while your embed snippet stays installed. Click "Activate agent" to turn it back on. It's handy for maintenance windows, seasonal closures, or while you're editing the agent.

No. The embed script loads asynchronously and defers initialization until it's needed, so it doesn't block your page render. Keeping customer sites' Core Web Vitals green is a hard requirement we test against — we run the same widget on this site.

On the Free plan the website widget shows a small "Powered by Dialogent" link in its footer; upgrading to any paid plan removes it automatically. The hosted chat page carries the same footer on all plans. Full white-labeling beyond that isn't self-serve yet — if it matters for your business (agencies especially), contact us about an Enterprise arrangement.

Yes. When a visitor asks for a human — or your escalation rules trigger — the agent immediately stops answering, collects contact details, and marks the conversation as escalated so your team follows up. Dialogent doesn't include a live-agent inbox: handoff means a fast, flagged follow-up by your team, not a live takeover.

Not today. Dialogent runs on your website (widget) and on a hosted chat page you can link from anywhere — including your Instagram bio or Google Business Profile. Messaging-channel integrations aren't built yet.

Yes — the widget runs on every page where the script is present, and the agent answers from your full knowledge base regardless of which page the visitor is on. Separate websites (or very different audiences) are better served by separate agents.

Free: $0 — 1 agent, 100 conversations/month. Starter: $29/month — 3 agents, 2,000 conversations, file uploads, webhooks, email support. Pro: $99/month — 10 agents, 10,000 conversations, the GEO site audit, priority support. Enterprise: custom — unlimited agents and conversations, SSO. No per-seat fees and no per-message credits.

A report on how citable your website is for AI search engines (ChatGPT, Perplexity and similar): it crawls up to 20 pages from your sitemap and scores each against citation-readiness checks — TL;DR blocks, question-format headings, FAQPage/Article structured data, image cadence and more — with a concrete fix suggestion for every gap. Run it from Dashboard → Analytics (Pro and Enterprise; one audit per hour).

One chat session: a visitor opens the chat and talks to your agent. Twenty messages back and forth in one visit is still one conversation. Page views don't count — a session is only created when someone actually opens the chat.

New conversations are paused until your limit resets or you upgrade — existing data, leads, and transcripts are never touched, and there are no surprise overage charges. You can see current usage any time in Settings → Billing.

At the start of each calendar month.

Better — a free plan with no expiry and no credit card required: 1 agent and 100 conversations a month, enough to run a real agent on a small site indefinitely. Upgrade only when you need more agents or volume.

Yes. Billing is monthly through Stripe with no contracts — upgrade, downgrade, update your card, download invoices, or cancel self-serve from Settings → Billing.

The public chat endpoints are rate-limited to blunt abuse, and sessions only start when the chat is actually opened and used. If you ever see abusive traffic patterns affecting your count, contact support and we'll sort it out.

Yes — run one workspace per client under a single login (each with its own agents, knowledge, integrations, and billing), or one agent per client inside a shared workspace (Pro covers 10 agents). Team roles cover account managers either way. A formal partner program with reseller pricing and white-labeling isn't launched yet; if you're managing multiple clients, contact us and we'll set you up.

Data lives in a managed Postgres database on our cloud infrastructure (DigitalOcean), reached only over TLS. Integration credentials are encrypted at rest (AES-256-GCM), every workspace's data is strictly isolated, public endpoints are rate-limited and input-validated, and all traffic is HTTPS.

No. We don't train models on your data. Conversations are processed by our LLM provider (OpenAI) via API to generate replies, and API data is not used to train their models either.

Our privacy policy explains exactly what we collect and why, including a section your business can point visitors to, and we honor deletion requests. A formal DPA and subprocessor page are in progress — contact us if you need one for your records.

No — Dialogent does not sign BAAs and must not be used to collect or handle protected health information. Clinics use it for what doesn't need PHI: hours, insurance accepted, service explanations, and booking. Configure your agent to redirect anything medical to your staff.

Every incoming message passes prompt-injection, relevance, and safety guards before the agent responds, and every outgoing answer is checked for grounding in your content. No system makes this impossible, but off-topic and manipulative inputs get refused — and full transcripts mean you can audit anything the agent said.

You control that. The agent answers only from content you provide, playbook rules let you require disclaimers and forbid advice-giving, and "talk to a human" triggers an instant handoff. Firms typically scope their agent to intake — case type, urgency, contact details — and leave the lawyering to lawyers.

Yes — deletion requests (yours or your visitors') are honored per our privacy policy. Contact us with the details and we'll remove the conversation data and any per-visitor memory.

Not yet — Dialogent is a young product and a formal audit hasn't been completed. Our concrete security measures (encryption at rest, tenant isolation, rate limiting, audit logging) are listed above and in the privacy policy; ask us for specifics your security review needs.

Yes. Invite teammates to your workspace as admin (everything, including billing), editor (manage agents and content), or viewer (read-only access to conversations and analytics) — so the front desk can read transcripts without being able to change the agent. Whoever creates the workspace is its Owner: an admin-level role that can't be changed, removed, or reassigned, so a co-admin can never lock the creator out.

Plans are priced by agents and conversations, not seats — invite the people you need at no extra cost.

Yes. Every workspace change — agent edits, member changes, settings — is recorded in an audit log under Settings → Activity.

Members of your workspace, per their role — the owner, admins, editors, and viewers can all read sessions. Nobody outside your workspace can access them.

Yes. Create as many workspaces as you need from the workspace menu (or Settings) and switch between them with one login. Each workspace is fully separated — its own agents, knowledge, integrations, team members, and its own plan and billing — so separate businesses or clients never share data.

Four usual causes: the agent is paused (check for "Activate agent" in the agent's menu), the script tag isn't actually in the published page (check your builder saved it), your site cache is serving the old page, or the domain doesn't match your widget settings. The dashboard's widget tab shows a live "installation verified" status the moment the widget first loads on your site.

Find the conversation in Sessions to see exactly what was said and which source it cited, fix or re-sync that source (or add an explicit Q&A pair for the correct answer), then verify with the Knowledge search tool. The fix takes effect within minutes.

The widget never breaks or errors at the visitor. If the engine is unreachable, the agent returns a safe fallback message inviting the visitor to leave contact details, so you still capture the lead.

Yes — every conversation is in the Sessions dashboard with its full transcript, outcome, and captured lead profile. Analytics aggregate the patterns: conversion funnel, top objections raised, and the questions your agent couldn't answer.

Dialogent runs on OpenAI models through our managed engine, wrapped with retrieval from your content, guardrails, and escalation logic. We select and upgrade models centrally — you never manage API keys or model settings.

Free plans have community support, Starter adds email support, and Pro gets priority responses. Whatever the plan, the fastest first stop for "why did it answer that?" is the session transcript — it shows the exact sources behind every reply.