HelpStackDocs

Guides

Channels

A channel is a connection between HelpStack and one of your communication platforms — an email account, a Facebook Page, an Instagram account, a WhatsApp number, or a website chat widget. Once a channel is connected, messages from that platform flow into your inbox and your replies go back out through it. This guide covers the channel types you can connect yourself, exactly what you enter to connect each one, the edit-channel settings, language and translation behavior, AI configuration, and how to read a channel card.

Who can manage channels? Connecting, editing, and deleting channels requires the OWNER or ADMIN role. Agents and viewers can see channels but not change them. See the team and roles guide.

Sections#

Adding a channel#

Go to Settings → Channels and start the add-channel stepper. It has four steps:

  1. Select type — choose EMAIL, FACEBOOK, INSTAGRAM, WHATSAPP, or WEBSITE_CHAT.
  2. Connection config — enter the connection details for that type (this differs per type; see below).
  3. Language + translation — pick the channel's language and set up auto-translation.
  4. AI settings — choose the AI provider, system prompt, knowledge base, and auto-reply mode.

Each channel type's step 2 is different, so the sections below focus on what you enter or click to connect each one.

Email#

To connect an email channel you provide both your outgoing (SMTP) and incoming (IMAP) server details:

  • SMTP (outgoing) — host, port, secure (on/off), username, password, and the from address used on messages you send.
  • IMAP (incoming) — host, port, and secure (on/off).

A Test button validates the connection so you can confirm the credentials work before saving. Once connected:

  • Inbound messages arrive via IMAP polling — HelpStack checks the mailbox on a schedule and pulls in new mail.
  • Outbound replies are sent via SMTP.

If you're unsure of your server settings, ask your email provider for their SMTP and IMAP details. Most providers publish them.

Ignoring senders#

Sometimes you don't want mail from a specific address, or a whole company, landing in your inbox — a noisy automated sender, spam that slips through, or a domain you simply don't support. An email channel can ignore specific senders so their mail never creates a conversation.

Add an entry two ways:

  • From the channel's connection settings (Settings → Channels → (your email channel) → Connection), under Ignored senders. Add a single address (jan@acme.com) or an entire domain (acme.com, which also covers its subdomains), then save.
  • From an email conversation, open the ⋮ menu and choose Add to ignore list to block that customer's exact address in one click.

Ignoring only affects new incoming mail — it doesn't touch conversations that already exist. If you've already been talking to that sender, the existing conversation stays right where it is; only future mail from them is dropped silently.

Email greeting & signature#

An email channel can add the greeting at the top and the sign-off at the bottom of every reply for you, so the AI only writes the message itself and every email ends with the name of the person who actually sent it.

Open Settings → Channels → (your email channel) → General and find Email greeting & signature:

  • Add greeting and signature turns it on. When it's off, replies go out exactly as before. New email channels start with it on; channels you connected earlier stay off until you switch it on.
  • Start with a greeting can be turned off if you only want the sign-off.
  • Greeting and closing per language. Each row is a language. Leave a field empty to use the suggestion shown in it (for example "Pozdravljeni," and "Lep pozdrav," in Slovenian, "Hello," and "Kind regards," in English), type your own, or type a single - to leave that phrase out in that language. The reply's language picks the row, so a customer who writes in Croatian gets "Poštovani," and "Srdačan pozdrav,". If there is no row for that language, the channel's own language is used, then English. These phrases are never machine-translated. Use Add language for a language that isn't listed.
  • Team name signs replies that nobody approved and nobody is assigned to, such as replies the AI sends on its own. It defaults to your workspace name.
  • Footer is optional small print at the very end, under a thin line: company name, address, phone or website.
  • Signature image is an optional logo or banner shown under the signature, above the footer. Upload a PNG, JPG, GIF or WebP up to 5 MB, set its width (up to 300 px) and, if you like, a link it opens. It is stored at a public address, because email apps load it when the customer opens the email. The plain-text version of the email has no image.

Who signs the email. The person who approves or sends the reply signs it. If nobody did (an automatic reply), the agent the conversation is assigned to signs it. If nobody is assigned either, the team name does.

My email signature. Each person sets their own signature on the Profile page, under My email signature: a name (your display name by default) and an optional second line (your workspace name by default, or none). It applies to the workspace you are in, so you can sign differently in each workspace you belong to.

Your own signature image. If you sign with a banner in your mail program (your name, phone, address and any disclaimer in one image), upload it under My email signature → Signature image and set its width (up to 600 px). Emails you sign then end with the closing and your banner, without the name lines, the channel's signature image or the footer, because the banner already shows them. The plain-text version of the email still has your name. An admin can also set it for a colleague under Settings → Team, by editing that member.

While a draft waits for approval, the conversation shows the whole email as it will be sent, with you as the signer. Editing the draft changes only the message in the middle. The editor opens in place of the message, grows with the text, and the expand button next to it opens a full-screen editor for long replies.

Facebook#

Connecting Facebook uses Meta's secure sign-in — you don't enter passwords into HelpStack:

  1. Click Connect / Continue with Facebook.
  2. Complete the Meta OAuth sign-in.
  3. Pick the Page you want to connect.

HelpStack then creates the channel with managed access tokens, so message delivery is handled for you. If a token is close to expiring, the channel card shows a token-expiry alert, and if the webhook needs attention a Fix Webhook action appears on the card so you can repair it in a click.

Instagram#

Instagram connects the same way as Facebook — through the Meta OAuth flow. Once connected, the channel handles your Instagram DMs. As with Facebook, the channel card surfaces token-expiry alerts and a Fix Webhook action when needed.

Recent Facebook and Instagram messages#

When you connect a Facebook Page or an Instagram account, HelpStack brings in your conversations from the last 30 days, so your inbox is not empty on day one. It takes a minute or two.

  • A conversation where the customer wrote last comes in open and unread, so you can see who is still waiting.
  • A conversation you already answered comes in closed. You can read it, and the customer's next message starts a new one.
  • Your own replies teach the AI how you answer, so its first drafts sound like you. This uses the Learn from past replies setting on the channel's AI & Knowledge tab.
  • A customer who wrote in the last 24 hours and is still waiting is treated like a new message. If the channel's AI reply mode is Auto Prepare, the AI writes a draft for you to approve (it uses credits like any new message: the draft, and the translation if auto-translate is on). On Auto Reply nothing is drafted or sent: connecting an account never sends a message.
  • The rest of the import sends nothing and uses no credits.

The channel's Connection tab shows how many conversations came in and how many are waiting for an answer. Import again picks up anything new without making duplicates. Photos and stickers show as [Media]. Meta shares the last 20 messages of each conversation, so a long conversation starts at its 20th newest message.

Facebook and Instagram only let a shop reply within 24 hours of the customer's last message, so older conversations are there to read.

WhatsApp#

To connect WhatsApp you use the Meta WhatsApp Cloud API. Enter:

  • Phone Number ID
  • Access Token
  • App Secret

HelpStack also shows webhook setup instructions so you can finish wiring the connection on Meta's side.

Website chat#

A website-chat channel (WEBSITE_CHAT) is the simplest to set up — there are no credentials to enter. When you create it, HelpStack generates a widget ID. You then copy the embed snippet (or the ID) and add it to your website so the chat bubble appears for your visitors. For the embed steps, see the widget embed guide.

Website-chat channels also let you configure per-channel FAQs and the widget's appearance right here in the channel's settings, so the widget matches your brand and answers common questions. See the FAQs guide.

Two of those appearance settings are pictures, and they are not the same picture:

  • Bubble icon — what the closed chat bubble shows on your site. Pick one of the built-in icons, or choose Your image and upload your own (JPG, PNG, GIF, WebP or SVG). Most people put their logo here.
  • Avatar — the face shown next to your replies inside the conversation. Usually a person, or your support team's mark.

You can set one, both, or neither. If you upload a bubble icon it is used on the launcher; if you have not, a custom launcher falls back to the avatar, which is how it behaved before the two were separated.

When you sign up and let HelpStack read your website, it also picks up your site's main colour and mark and applies them to the widget for you. Anything you have already chosen yourself is left alone.

Product cards#

When a reply in the website chat links to a product page on your shop, the visitor sees a card instead of a bare link: the picture, name, price and stock, with an Add to cart button right in the chat.

Product cards under a chat reply, and the same card after tapping Add to cart
Product cards under a chat reply, and the same card after tapping Add to cart

There is nothing to set up. HelpStack reads the product details from the linked page itself, from the data your shop already publishes for Google, so the card always shows what your shop says. Your AI can't make up a price, and a link that isn't a product (your shipping page, a blog post) simply stays a link. Cards work for replies from the AI and for replies your team writes.

  • Add to cart works out of the box on Shopify and WooCommerce. After a tap, the button changes to Added · View cart, and your shop's own cart updates.
  • A product with sizes or colours shows Choose options, which opens the product page.
  • A product that is sold out shows Out of stock, so nobody is sent to a dead end.
  • A reply shows at most three cards, side by side to swipe through on a phone.

When the visitor taps a card button, the AI knows on its next reply ("they added the pour-over set"), and the tap is recorded in your tool logs.

On another shop system, or to add your own button (Save for later, Book a fitting), your developer can register it on the page in a few lines; see Product cards in the widget guide.

To turn cards off, open Settings → Channels → (your website chat) → Appearance and switch off Product cards. Links then show as plain links.

Business hours & offline messages#

A website-chat widget can show visitors when you're actually available, and collect a message instead of leaving them stuck when nobody's on shift. Configure this under Settings → Channels → (your channel) → Appearance → Availability:

  • Business hours — set opening hours for each day of the week, plus a timezone. Outside those hours, the widget shows as offline and visitors can't start a live chat.
  • Temporarily closed — a manual override that shows the widget as offline right now, regardless of the schedule. Useful for a holiday or an unplanned outage.
  • Collect email when offline — when this is on, the message box is replaced by a short form (name optional, email and message required) any time the widget shows as offline, whether that's from your business hours or the Temporarily closed override. Submitting it creates a normal conversation in your inbox and shows the visitor a confirmation message.

Customize what visitors see while you're offline in the Offline Message field — it appears above the capture form (or on its own, if capture is off).

Since a visitor who leaves a message this way has usually already left your site, your team typically follows up by email — see Continue via email.

If business hours are off, the widget always shows as online and the offline form never appears.

Telegram (coming soon)#

Telegram exists in the system but is not yet available to connect from the interface. Support for it is coming soon — for now, choose one of the other channel types.

Channel types not shown in the picker#

Three more channel types exist in HelpStack but do not appear in the "Add channel" list by default, because each one needs to be switched on for your organization first:

  • Voice — an AI voice agent that answers spoken questions through an embeddable widget. Requires the voice feature to be enabled for your organization.
  • LinkedIn — LinkedIn messaging. Gated on LinkedIn partner approval, so it is off unless your organization has been enabled.
  • Telegram — see above.

If you want any of these, contact support and we will tell you whether it is available for your account. They are not self-serve today.

Editing a channel#

Open a channel to edit it. The edit dialog is organized into tabs:

Basic

  • Type — the channel's platform.
  • Name — a friendly label for your team.
  • Language — the channel's primary language.
  • Auto-translate — turn translation on or off for this channel.
  • Fallback language — the language to use when detection is uncertain.

Connection

  • The channel's credentials for its type.
  • Test / Reconnect — validate or re-establish the connection.
  • Tools — access to the channel's agent tools.

AI & Knowledge

Reachable directly at Settings → Channels → (your channel) → AI & Knowledge:

  • AI provider — which configured provider drafts replies for this channel.
  • System prompt — your instructions to the AI about tone, policy, and behavior.
  • Knowledge-base group assignment — which knowledge groups the AI can draw on.
  • Auto-reply mode — Off, Auto Prepare, or Auto Reply.

Language & translation#

Translation is configured per channel and is what lets you and your customers speak different languages seamlessly:

  • Auto-Translate toggle — when on, HelpStack translates messages in both directions for this channel.
  • Agent language — incoming customer messages are detected and translated into your (the agent's) language so you can read them. This comes from your preferred language; see getting started.
  • Outgoing replies — your replies are translated into the customer's language before sending.
  • Fallback language — used when the system can't confidently determine the language.
  • Short-message behavior — messages under six words inherit the conversation's already-detected language. This prevents the detected language from flip-flopping between similar languages on tiny messages like "ok" or "thanks".

In the inbox you can always override translation per message with Send Original and Retranslate — see the conversations guide.

AI & knowledge#

Each channel decides how the AI helps on it. In the AI & Knowledge tab you set:

  • AI provider — the OpenAI or Anthropic connection that powers replies. Configure providers in the AI providers guide.
  • System prompt — the standing instructions the AI follows on this channel.
  • Knowledge-base groups — the information the AI may reference; build these in the knowledge base guide.
  • Auto-reply mode — Off (manual only), Auto Prepare (AI drafts, you approve), or Auto Reply (AI drafts and sends). Auto Prepare is the default for new accounts. The full behavior is explained in the AI replies guide.

Reading the channel card#

Each channel appears as a card on Settings → Channels. At a glance the card shows:

  • Active / Inactive status.
  • Translation settings for the channel.
  • Assigned knowledge-base groups.
  • System-prompt preview.
  • AI provider in use.
  • AI reply mode (off / Auto Prepare / Auto Reply).
  • Email poller status (email channels) — last polled time and any poll errors.
  • Facebook/Instagram token-expiry alerts when a token needs renewing.
  • Webhook status, with the Fix Webhook action when repair is needed.
  • Widget ID (website chat) — with a copy button to grab it for your site.

Use the card as your health check: a glance tells you whether a channel is live, whether AI is active, and whether anything needs attention.

Related#

Edit this page on GitHub