Built for · Customer support
Clear the queue. Keep your voice.
Bash triages the queue and drafts every reply in your voice.
Nami knows every account and what was promised.
Bash · on its own machineWorking
YouDraft the reply to Deniz at Fernbrook
Read this thread and the two before it
Pull the account — tier, renewal date, open promises
Find what we told them in March about the export capDraft the reply in your tone, hold it for your wordHeld · waiting on you
One reply, draftedFernbrook: “You were right that we said the export cap would lift this quarter. It did, on the 12th. Your workspace was left on the old limit, which is on us, and I’m moving it this afternoon.”
Held for your wordWatch it work
WhatsAppTriage this morning’s queue↵DelegateRead the 23 tickets that landed in the shared inbox overnightMatch each sender to the account — tier, renewal, open promisesGroup them: down, blocked, question, thanks
Route the outage to #support-urgent with the log excerpt
One ask. The whole queue moves.
Pick a delegation — Bash runs it end to end, held before it reaches a customer.
WhatsAppTriage this morning’s queue↵DelegateRead the 23 tickets that landed in the shared inbox overnightMatch each sender to the account — tier, renewal, open promisesGroup them: down, blocked, question, thanksRoute the outage to #support-urgent with the log excerptTriage · posted in #support-urgent
23 in, 1 on fire
Halcyon Labs has been failing bulk imports since 04:10. Enterprise, 240 seats, renews in six weeks. That one is in #support-urgent with the log excerpt. The other 22 are grouped, 14 already drafted.
Done · 40 seconds
From wherever you areWhatsApp7:58 AM#supportMon 9:02GmailFwd
or press ⌃ ctrl anywhere on your Mac
Ask from the kitchen. The queue is sorted before the kettle.
Gmail, Slack, WhatsApp, iMessage, Telegram — or a key on your desk.
WhatsApp7:58 AMwhat came in overnight7:58 AM
Triaged. 23 overnight · 1 down. Halcyon Labs has been failing imports since 04:10 — that one is in #support-urgent. The other 22 are grouped, 14 already drafted.7:58 AM · 40 seconds
#supportMon 9:02@Bash build this week’s health digest9:02 AM
Posted in #support · the sheet is linked. 168 tickets, first response 41 minutes, three accounts at risk. Kestrel Freight renews in 30 days.9:03 AM
GmailFwdFwd: Re: import dropped half my rows — can you take this?2:14 PM
Reproduced it, filed #4412 with the logs and pinged #eng. Replies written for all three customers on it — held for you.2:17 PM
or press ⌃ ctrl anywhere on your MacSigned into your world
Works inside your stack, as you.
The shared inbox, the room, the issue tracker — and a thousand more, a click away.
GmailTriage the shared inbox, draft the reply, send on your word.
SlackEscalate the outage, answer the thread, post the digest.
SalesforcePull the tier, read the history, log what was promised.
GitHubFile the repro, watch the issue, carry the fix back to the ticket.
NotionFind the macro, update the help doc, retire the stale answer.
Google SheetsCount the volume, track first response, flag the accounts slipping.
ZoomJoin the call, watch the repro, write the recap.
WhatsAppSummon triage from anywhere — the queue sorts before you sit.
StripeRefunds and credits, calculated and held for your word.
— and 1,000+ moreSee everything it connects to
Docs
Teams
X
Googlecalendar
Linkedin
InstagramSay it like this
Delegate in plain words.
No setup, no syntax — it already knows your accounts, your macros, your tone.
Triage everything that came in overnight. Group by urgency, name the account and tier on each, and tell me who has waited longest.
Triage
Draft the reply to this thread using Fernbrook’s tier, their last three tickets and what we promised them in March. Sound like me. Hold it.
Reply
This is a real bug. Open a GitHub issue with the repro and the logs, ping #eng, and link the issue back on every ticket reporting it.
Escalate
Pull last week’s ticket volume by account and type into a sheet, flag anyone trending up inside 60 days of renewal, and post it to #support.
Health digest
Find every ticket older than 48 hours with no reply from us, tell me why each one stalled, and draft the apology on the worst three.
Stale queue
The import limit answer is wrong in three macros. Rewrite them from what shipped in issue #4412 and hold the changes for me.
Macros
Routines that fire themselves48 hours without a reply from us → an apology drafted in your tone, with the reason it stalled, held for your word.Gmail → draft · held
On a schedule, or on a signal.
Set it once. You hear about it when it’s done — and only then.
Daily7:30
Overnight tickets grouped by urgency, each carrying the account and tier. Anything down is already in #support-urgent.ScheduledQueue triaged before you sit down
Mon9:00
Volume by account, first response, who is trending up before a renewal. Sheet built, the read posted to #support.ScheduledCustomer-health digest
TriggeredA ticket goes cold
Gmail → draft · heldFri16:30
Every open issue a customer is waiting on. Status pulled from GitHub, and the ones that shipped come back as drafted replies.ScheduledEscalations, reconciled
Nami · the keeperOldest is Kestrel Freight · 3 days, 2 hoursEnterprise · 240 seats · renews Oct 9GitHub · merged 22 minutes ago
It remembers every customer. So nobody explains themselves twice.
Tier, history, what you promised — filed once, and noticed when it changes.
Nami · entity
Halcyon Labs
AccountEnterprise · 240 seats · renews Oct 9
ContactAna Beck, Head of Ops · replies within the hour, hates being handed to someone new
History9 tickets this quarter · 6 of them bulk import
Promised“The 5,000-row cap goes away in Q3” — ticket 8871, Mar 18
OpenIssue #4412 · filed Tuesday, no ship date yet
Kept since her first ticketNoticed on its own
Four tickets have gone 48 hours with no reply from us. Draft the apologies?
Oldest is Kestrel Freight · 3 days, 2 hoursNot nowHandle it
Noticed on its own
Halcyon Labs has filed six import tickets this quarter and renews in six weeks. Brief the account team?
Enterprise · 240 seats · renews Oct 9Not nowHandle it
Caught a change
Issue #4412 shipped this morning. Reply to the three customers waiting on it?
GitHub · merged 22 minutes agoNot nowHandle it
Asks before it matters
Nothing reaches a customer unread.
Every reply, anything public, any credit or refund — held until you say go.
Held · waiting on you
Send the reply to Deniz at Fernbrook?
To deniz@fernbrook.coRe Re: CSV export still capped at 5,000
You were right that we said the cap would lift this quarter. It did, on the 12th — your workspace was left on the old limit, which is on us. I am moving it this afternoon and will write back when it is done…
EditHoldApprove & sendGmail · as you
Only the tools you connectGmail, Slack, Salesforce, GitHub — nothing else. Disconnect any of them, any time.
Every step, in the openThe triage, each tool call, the draft — visible while it works. Peek in, or take the wheel.
Your customers stay yoursTickets, accounts and history live with you, not with a model. Switch minds; the work comes along.
Taught once, done your way
SupportQueue and replies. Triage by urgency, the account in front of every draft, your tone, nothing sent unread.TriageRepliesFollow-upTone
EscalationsThe real bugs, filed properly. Repro, logs, the accounts affected, and the issue linked back on every ticket waiting.ReproIssuesStatus
Create your ownYour playbook, encoded. Teach it what you escalate, how you apologize, when you refund — in plain words.Your rulesYour voice
Install the Support playbook. Or teach it yours.
An Ability is everything the job needs — bundled, installable in a click.
SupportQueue and replies. Triage by urgency, the account in front of every draft, your tone, nothing sent unread.TriageRepliesFollow-upToneEscalationsThe real bugs, filed properly. Repro, logs, the accounts affected, and the issue linked back on every ticket waiting.ReproIssuesStatusCreate your ownYour playbook, encoded. Teach it what you escalate, how you apologize, when you refund — in plain words.Your rulesYour voiceCustomer support · FAQ
Questions, answered.
Does it reply to customers on its own?
Not unless you tell it to. Every outbound reply is drafted and held for your word, with the tier, the history and what you promised sitting next to it. Triage, research, internal notes and issue filing keep moving. You can loosen that for the routines you trust, one at a time.
Will it sound like our team, or like a bot?
It learns from the replies your team has already sent — the length, the openers, how you apologize, what you never promise. Nami keeps the rules you state once ("never blame the customer", "always give a date"). You’ll edit the first few; after that, mostly you just approve.
How does it know a ticket is a real bug?
By checking, not guessing. It looks for the same symptom across the queue, reproduces it where it can, and reads what engineering already has open. One report becomes a question for you. Three reports with one repro become a GitHub issue, an #eng ping, and a link back on every ticket waiting on it.
What does it know about our customers on day one?
Whatever’s in the tools you connect: the shared inbox, the account record, the issue tracker, last quarter’s threads. Helpdesks connect the same way: Zendesk, Intercom, Front, Freshdesk. Their history comes with them. From there Nami keeps building: who complained about what, and what was promised.
Where does the work run?
On its own computer in the cloud — a real machine with a browser, files and a terminal — or on yours when the job lives there. Its screen streams as it works; you can peek in or take the wheel.
The door
Come back to answered.
Send us a week of your queue. We’ll show you what Bash would answer, what Nami would keep, and what’s actually left for you.