← Back to library

Sell a service through a chat agent that checks payment

degutemesgen on Hacker News describes sales agents that move through a fixed lifecycle of discovery, proposal, checkout, delivery and completion. Payment happens through a Stripe link and must verify before any work is handed over; refunds and disputes route to a person by chat.

  • OutcomeOffer a simple service through a messaging agent that moves from inquiry to proposal to paid checkout to delivery, and pings you before refunds or disputes.
  • SetupAn afternoon+
  • LevelAdvanced
  • Creditdegutemesgen on Hacker News
View original
Works with
Custom agent

Published Oct 23, 2026Updated Oct 23, 2026AI dry run on sample input: passed Oct 23, 2026Source verified Oct 24, 2026

When should you use it?

  • You sell a simple productized service
  • Customers ask questions before buying
  • You want refunds approved by a person

Not for: Complex custom quotes; Regulated services needing licensed advice

What do you need, and what do you get?

You need

  • Service description and price
  • Stripe account
  • Owner's Telegram

You get

  • Proposals in chat
  • Paid orders with delivery
  • Approval requests for refunds

How does it work, step by step?

  1. Define the stagesWrite what the agent does at each stage from first question to completion.
  2. Generate checkout linksCreate a payment link for the agreed offer.
  3. Verify before deliveringConfirm the payment succeeded before sending anything.
  4. Escalate money questionsSend refunds and disputes to the owner and wait for a reply.
  5. Review conversationsRead transcripts weekly and refine the offer text.

The prompt

Our own prompt, inspired by degutemesgen on Hacker News and written to work in Custom agent. Values in lime are placeholders.

Fill in the blanks (optional; the copied prompt uses your values)

Prompt3 variables
You sell {service} for {price} on {channel}.
Stages: explore the need, propose the fixed offer, send a checkout link, confirm payment succeeded, deliver, confirm completion. Never change price or scope. For refunds, disputes or anything off-script, message the owner and wait for approval.

Inspired by degutemesgen on Hacker News: Hacker News post describing an agent that sells services over chat. The prompt above is written by playbooks.bot in our own words; nothing from the original is quoted.

What can go wrong?

  • Delivering on a checkout link instead of a confirmed payment invites fraud.
  • Agents can promise extras; keep the offer text fixed.

Our take

The fixed lifecycle is what makes this safe enough to try: the agent always knows which stage it is in, and money moves only through a checkout link and a verified payment. Routing refunds and disputes to a person by chat is the right call. Separate sandbox and live keys. Write down what the agent does at each stage, from first question to completion, and keep the offer text fixed so it cannot promise extras. Generate a checkout link for the agreed offer and confirm the payment actually succeeded before delivering anything; delivering on a link alone invites fraud. Refunds and disputes go to the owner, and the agent waits for a reply. Read transcripts weekly to refine the offer wording. Keep test and live payment keys separate. Complex custom quotes and regulated advice are not a fit. The fixed stages also make transcripts easy to review, because you can see exactly where a conversation stalled or went off script.

Where the time goes back: Routine inquiries turn into paid orders without you in every chat, and you only handle exceptions.

FAQ

Who handles refunds?
The owner, approving by reply in Telegram.
Does it need hosting?
The author says it runs locally and connects to messaging apps or a widget.

Source and credit

Inspired by degutemesgen on Hacker News. We describe the idea in our own words, link to the original, and wrote the prompt ourselves; nothing is quoted.

Inspired by
Hacker News post describing an agent that sells services over chat
Creator
degutemesgen on Hacker News
Published on
Hacker News
Original date
May 9, 2026
License
Not openly licensed (the poster keeps all rights), so we quote nothing and link to the original
Originally for
Custom agent
AI dry run
Oct 23, 2026: passed. Dry-run tested by our AI test agent on sample input. PASS: Gates held on desk check; no live sends or purchases.
AI dry run on sample data; not run with live connected accounts. The playbooks.bot test agent filled the prompt with synthetic inputs, produced the output and checked it against the page's stated outputs and guardrails.
Verified
Oct 24, 2026: source link resolves to the original, the credited creator published it, our excerpt or prompt matches it, and the license allows our use.

Are you the creator? Claim or correct this page.

More in Sales

All 59 Sales playbooks →

Related playbooks

SalesWorkflow✓ AI dry run

Draft researched cold-email openers from each site

Make cold outreach feel researched by having a workflow read each lead's own website and draft personal openers, while you stay honest about who did it.

Credit: James Presbitero

An afternoon+
SalesWorkflow✓ AI dry run

Qualify web leads with AI that must quote its evidence

Route every website inquiry through a workflow that removes duplicates and junk, has a model judge each of your qualification rules with a supporting quote from the message, uses plain code to confirm each quote really appears in the inquiry, updates the CRM, drafts a reply that asks only for missing details, and sends nothing until a person approves it in chat.

Credit: Kingsley_Ken

An afternoon+
SalesWorkflow✓ AI dry run

Run a real estate deal desk from your inbox and calendar

Give an agent your inbox and calendar so it surfaces counteroffers fast, compares competing bids, drafts routine deal emails and warns you about clashing deadlines.

Credit: Drew McAllister (@DMAC_19)

An afternoon+