Venom Support
One shared inbox for customer email and form requests in Venom Cloud. Assign, snooze, tag and close conversations, write internal notes with @mentions, use saved replies, draft replies from your notes, and optionally let fully covered drafts send themselves after a review delay.
Venom Support is a shared inbox for your customers, at /support (g s from anywhere in Venom Cloud). Mail your
customers send and requests from your forms land in one list. Each conversation has an owner,
a status, a priority, tags and a clock, and a reply is drafted for you from your own notes.
Everyone in the org sees every conversation: that is what makes it shared. The one exception is a conversation assigned to a private team, which only that team's members and the org's owners and admins see. Owners, admins and developers can reply and change things; billing and viewer roles read.
The inbox
The list has six views with counts: Mine (open, assigned to you), My teams (open, assigned to a team you are in,
subteams of your teams included), Unassigned (no person and no team), All open, Snoozed and Closed. Search (/) looks in subjects, customers and message text, and #42 finds a number. A row shows the
customer, the subject, the last words, the priority, an SLA badge when a deadline is near or missed, the first tags and
the owner. The list refreshes on its own every 30 seconds.
| Key | Does |
|---|---|
j / k |
Next, previous conversation |
Enter |
Open it (on a phone) |
e |
Close it, or reopen a closed one |
a |
Assign (type to find a teammate, or unassign) |
s |
Snooze (1 hour, 3 hours, tomorrow 9:00, next Monday, a week, or 45m, 2d, a date) |
r / n |
Write a reply / an internal note |
c |
New conversation |
/ |
Search |
Tab |
Next view |
Esc |
Back to the list |
A conversation can also be assigned to an org team, with or without a person: pick it with the Team
button next to Assign. Whoever joins the team sees it under My teams right away, and whoever leaves stops seeing it
there. Over MCP, support_conversation_assign takes team (a key or id, null to clear) as well as assignee, and
support_conversation_list takes view: "teams" and team.
Closing, snoozing or assigning away moves you to the next conversation. A snoozed conversation returns to Open at its time, or sooner when the customer writes; a closed one reopens when the customer writes again.
Replying
The composer under a conversation has two tabs. Reply goes to the customer, with the thread's subject and mail
headers, copying the people on the thread. Internal note is for your team and is never sent; write @ana (the part
of a member's email before the @) and Ana gets a bell notification until she opens the thread. ⌘↵ sends.
Send and close does both. Saved replies inserts a template with the customer's name and yours filled in
(manage them in Settings).
Replies leave only through an email provider you connect (your SMTP server, or your own Postmark, SendGrid, Mailgun or Resend key: Support email). Venom has no mail account of its own for you. With none connected, a reply is still kept on the thread but marked not sent, with the reason and a link to connect one: it is never dropped silently.
A conversation that came in on WhatsApp, SMS, RCS or Telegram is answered through Venom Comms on that channel instead, from the number the customer wrote to, and shows ticks (sending, sent, delivered, read). Outside WhatsApp's 24-hour window the composer offers approved templates, or sends by SMS when your support profile allows it. Files the customer sent show in the thread. All of it is in Support and People.
Drafts from your notes
Draft with AI searches the notes in your org vault in Notes for the subject and the customer's last
message, and writes a reply from them with your org's own gear. The draft appears on the conversation with the notes it
used as links; if no note matched it says so, so you read it harder. You can Approve and send it, Edit it in the
composer, or discard it. Nothing is sent without a person pressing send, unless your org turned on automatic replies
(below). Internal notes inform a draft but are never quoted in it. Drafts count against a daily AI limit per org
(default $1; support_ai_daily_cap_usd in the org's settings), and say so when it is reached.
Automatic replies
Support can answer a customer by itself, narrowly, when your notes fully answer the question. It is off until an owner
or admin turns it on in Settings, Automatic replies, and only an owner or admin changes it there (an agent reads it
with support_autoreply_get but cannot turn it on).
What it sends. When a customer writes on a channel you allowed (email, chat, SMS, WhatsApp, RCS or Telegram), Support writes the same draft Draft with AI would, in the background. The model also says whether the cited notes fully answer the customer's last message. Only when it says yes, and at least one cited note is a real note in your org vault, does the draft get a send time: it waits on the conversation for the review delay (default 10 minutes, 0 to 1,440), marked "Sends by itself in 8 minutes". Until then anyone who can reply can Send now, Edit it (the timer stays), Hold it (the timer goes and it waits for a person) or Discard it. At the time, it goes out through the normal reply path: your email provider, or Comms on the customer's channel, with the same rules as a reply you send. The thread shows it as Sent automatically, by "Venom Support (automatic reply)", with the notes it cited. The customer sees the reply text and nothing else.
Every other draft stays an ordinary draft for a person, exactly as without the setting: the model said the notes do not fully answer the question, gave no readable verdict, cited nothing, or cited a note that no longer exists.
Which conversations. The channels you tick, and optionally only conversations with one of the tags you list (a messaging number's inbox is a tag) or assigned to one of the teams you tick.
What it never does. It never replies when a member already replied in the conversation (a reply from the WhatsApp
Business app counts), the assignee wrote in it in the last day or a member's own draft is pending, the conversation is
closed, snoozed or tagged spam, the message is an auto-responder or a bounce (out of office, Auto-Submitted, mailer
daemons), or the customer asks for a person ("can I talk to a human", "real person", "are you a bot", "call me"). It
never closes a conversation, never sends a file or a WhatsApp template (outside WhatsApp's 24-hour window it goes by SMS
only when your support profile allows that, as any reply would; otherwise the draft waits for a person and a note says
why), and never touches Outreach: a reply to a sequence still stops
the sequence. A new customer message stops a draft that was still waiting; it stays as an ordinary draft. Turning the
setting off stops every waiting draft.
Limits. At most 1 automatic reply per conversation (up to 5), so a customer who writes again after one gets a
person; at most 50 a day for the org (UTC; up to 1,000); and the drafts count against Support's daily AI limit
(support_ai_daily_cap_usd), which stops them when it is reached. The settings show today's count. Every change to the
setting is in the audit log as support.autoreply.update, and every automatic reply as
support.autoreply.sent (ids and counts, never the text). An automatic reply does not count as an approved AI reply on
the bill (below).
Agents: support_draft_update changes a draft's text or holds it, support_reply_send with use_draft sends it now,
and support_autoreply_get (GET /api/v1/support/auto-reply) lists the settings, today's count and the drafts waiting
to send with their send time.
Customers and People
A customer is whoever wrote: an email address (or, on a messaging channel, a phone number or Telegram chat) and a name, taken from the message. If the same email or phone number belongs to someone in People that your org shares (visibility "org"), the conversation links to them. Private and team-only people are never read by Support.
Issues, errors and ratings
Create issue (the button under the conversation's title, or support_issue_create) makes a Tasks issue in a team you choose. It carries the last messages of the thread (internal notes stay out) and a link back, lands in the team's triage, and is linked to the conversation, which then shows the issue's identifier and state. The link back needs the cloud's address: it is the one you are browsing, or VENOM_CLOUD_ORIGIN when an agent makes the issue. When the linked issue is completed, the next time someone opens the conversation a "we fixed it" draft is waiting for you to read, edit and send. It is prepared once per link and never sent on its own; turn it off in Settings, After a conversation. If the issue is in a private team you're not in, the conversation shows a Private issue chip, with no key, title or status. The chip links to the issue, so the team's members can open it. Support's notes don't name the issue.
Add to issue (next to Create issue, or support_issue_link) adds the conversation to an existing issue as a customer request, with an optional quote, so the issue counts who asked for it. A conversation with no issue yet is linked to that one, and the customer still gets the "we fixed it" draft when it is completed. Create issue makes the conversation the new issue's first request.
Errors seen by this customer. Set the customer's user id (customer_user_id on support_conversation_update) and the conversation lists the errors Incidents saw in the last 30 days for a user with that id or with the customer's email address (the user context your SDK sends). Each links to its error issue.
CSAT, a rating when a conversation closes. Off by default; turn it on in Settings, After a conversation. When a conversation closes, the customer gets one plain-text email through the email provider you connected, with two links, good and not good. The links point to a page on your own Venom Cloud, /csat/{token}, signed with a key only your org has and impossible to guess; opening the page records nothing (mail scanners open links), the customer presses Send, and can add a comment and change their mind. There is no tracking pixel and nothing else is contacted. A conversation is asked once, and not at all when the customer has no email address or no provider is connected (a note on the thread says why). The rating and comment show on the conversation, are in a note on the thread, and agents read them with support_conversation_get. The rating form is limited to 20 requests a minute per client.
Chat on your site. Visitors can chat from a widget on your website and the conversations arrive here on the channel chat; see Chat widget and help center.
Webhook. The org webhook events support.sla.breached (see SLA policies and reports) and support.conversation.created (Settings, Webhooks) fires for every new conversation: id, number, subject, channel and priority, never the customer's message.
Where conversations come from
- Email: point your provider's inbound webhook at Support (Support email).
- Forms: add the destination Support conversation to a form (Forms destinations).
- WhatsApp, SMS, RCS and Telegram: route a Comms number to Support in Settings, Messaging channels; each customer who writes gets a conversation, answered on the same channel (Support and People).
- By hand:
c, for a call or a chat. - Agents: Build on Venom Support.
The bell shows @mentions, your conversations waiting on a customer's last word, conversations past their SLA and the number nobody owns.
Service levels
Each conversation gets a first-reply deadline and a resolution deadline from two org-wide targets (default 4 hours and 2 days), scaled by priority: urgent a quarter, high half, low double. The badge turns amber in the last hour and red when missed. A reply that was actually sent counts as the first reply; one saved as not sent does not.
AI-resolved conversations on the bill
When you send an AI draft as written (Send the draft, without editing it), Support remembers that the last reply on the conversation came from the draft. If the conversation then stays closed for 7 days without being reopened, it counts once as a "Support conversation resolved by an approved AI reply" on the outcome meter (Billing and usage). A reply a person wrote or edited takes the credit away, and a reopened conversation starts its 7 days again when it is closed. Each conversation is counted at most once.