Guide

How Sloth-Butler works

How to use it, what's under the hood, common questions and what's next.

01

What it is

Sloth-Butler handles the messages you dread replying to: a client asking for free extra work, a customer pushing for a discount, an invoice that is weeks late, someone who wants to pay later, or a last-minute call. You paste the message, and it writes a polite, firm reply with one counter-offer, such as a paid option, a later date or a partial yes. You can copy it, listen to it or save it as a voice note, then log the hours and money it protected. It remembers your rate and your floor, and plans follow-ups for invoices that are late.

Write a reply
02

How to use it

The labels below match the buttons on the main page. It takes about a minute.

  1. 1

    Save your terms once

    In “Your terms”, add your name or business, currency, hourly or day rate, minimum project price, the biggest discount you’ll give, payment terms, late fee and when you’re next free. Press “Save my terms”. They stay in your browser.

  2. 2

    Pick the situation

    Free extra work, a discount request, a late or unpaid invoice, “I’ll get back to you”, a pay-later request, a last-minute favour, or Something else. This tells the model what kind of counter-offer makes sense.

  3. 3

    Paste their message

    Copy the email, DM or text you received into “Their message”. “Use an example” fills in a sample if you just want to see how it works.

  4. 4

    Add the amount or deadline

    Both optional. The amount is used in follow-ups and in your log; the deadline lets the reply name a real date.

  5. 5

    Choose a tone and language

    Soft is warm, firm is short and direct, maybe-later names a time to revisit. Languages: English, Spanish, French, Portuguese, German and Nigerian Pidgin.

  6. 6

    Press “Write my reply”

    You get one reply with a single counter-offer that uses your real figures. “Using your terms” shows it read them. If a draft quotes less than your minimum or more discount than you allow, it is rewritten or a line with your real terms is added.

  7. 7

    Plan follow-ups

    For a late invoice or “I’ll get back to you”, press “Plan follow-ups”. You get a friendly nudge, a firmer reminder and a final notice with your late fee, due in 3, 7 and 14 days unless you change the numbers. They land on the Follow-ups board under Due today, Upcoming and Done. The number next to Follow-ups in the menu is what’s due today.

  8. 8

    Copy, play or download

    “Copy reply” puts it on your clipboard. “Play voice” reads it in the ElevenLabs voice Alice. “Download mp3” saves it as a voice note you can send on WhatsApp, Telegram or anywhere else.

  9. 9

    Mark sent, mark paid

    Copy a follow-up or play it as a voice note, send it yourself, then press “Mark sent”. When the invoice is paid, “Mark paid” closes the plan and adds the amount to your log as money recovered.

  10. 10

    Log what it protected

    If you send a reply, enter the hours and money it saved you and press “Save to log”. Totals count up on the page.

  11. 11

    Check the proof

    Lower on the page, compare the same message on plain Qwen3-8B and on Sloth-Butler, and read the test scores.

Messages to try

Copy one and paste it into “Their message”.

Free extra work

Since you're already doing the logo, could you also put together a few social media banners? Should be quick for you.

Discount

Love your work but $1,200 is more than we planned. Could you do it for $700?

Late invoice

Hi, sorry for the delay on invoice #104. Things are slow this month, can we settle it in six weeks?

Pay later

Can you ship the two lamps today? I'll transfer the money once my client pays me.

Last-minute

Any chance you could jump on a call in 20 minutes? Just want to run some ideas past you.

03

How it works

The model is Qwen3-8B, an open-weight model anyone can download.

I trained a small add-on for it (a LoRA adapter, rank 32) on Tinker using 1,101 example replies to client, customer and invoice messages in six languages, over 204 training steps. Every reply is polite, firm and makes one concrete counter-offer. To test it, an AI judge scored 100 messages it never saw: a pass means at least 4/5. Replies that pass all five checks went from 41% to 63%, and they got shorter (51 to 34 words on average). The situation you pick and your saved terms are added to the instructions, so the counter-offer fits the message.

Before training means plain Qwen3-8B with the same instructions. After training is the same model with the adapter on. The compare button on the main page shows both.

Voice is ElevenLabs, voice Alice. It only runs when you press Play or Download mp3.

Their messageplus situation, your terms, tone
Qwen3-8B + adapterruns on Tinker
Replyfirm answer, one offer
ElevenLabs (Alice)play or download mp3
04

FAQ

  • It remembers your terms. Your rate, minimum, discount limit and payment terms go into every reply, and the server checks the reply doesn't go under your floor.
  • It follows up for you. A late invoice gets a three-step plan with dates, and the board tells you what to send today.
  • It's a small open model trained for this one job. Qwen3-8B with a small adapter, MIT licensed. You can run it yourself.
  • Your messages stay private. No account, no database, and message text isn't logged. Terms, follow-ups and your log live in your browser.
  • Voice notes. Any reply or follow-up can be played or saved as an mp3 to send on WhatsApp or Telegram.

The server reads every reply for prices and discounts. If it quotes less than your minimum project price or more off than you allow, it writes the reply again with that pointed out. If the second try still breaks it, it adds a line with your real terms, such as “To be clear, my minimum for a project is $900.” Hourly rates, deposits and part payments aren't counted as the project price.

For the messages that cost you time and money: requests for free extra work, pressure to drop your price, invoices that are late, and people who go quiet on a decision. If you bill for your time or your products, it helps.

No. There is no database. Your message and context go to the model on Tinker to write the reply, and the reply goes to ElevenLabs only if you press Play or Download. The server logs service errors, not your text. Your terms, follow-ups and log of protected hours and money live in your browser's local storage.

Two or three options turn the reply into a negotiation. One clear option, like a paid add-on or a fixed payment date, keeps the conversation short and keeps your terms in place.

You enter it. For a discount request it might be the discount you didn't give; for scope creep, the value of the work you didn't do for free; for a late invoice, the amount you asked for. Hours work the same way. The app adds up what you log, in the currency you choose.

The log supports USD (default), EUR, GBP, CAD, AUD, INR, JPY, BRL, ZAR, MXN, CHF and SGD. Replies can be written in English, Spanish, French, Portuguese, German or Nigerian Pidgin. All six languages were part of training.

No. It writes the reply. You read it, copy it or download the voice note, and send it yourself.

Free to use. The code, training data and eval scripts are MIT licensed: github.com/comzzy/sloth-butler.

  • Replies can be formulaic, often “I can't do X. I can do Y.”
  • It doesn't know your contract or your real calendar. Check the counter-offer is something you'll actually honour.
  • The test scores come from one AI judge, not people. On 100 held-out messages, 63% of trained replies passed all five checks, up from 41% before training. That still leaves plenty of misses.
  • Pidgin improved (10% to 50% judged natural) but only 10 replies were tested, and no native speaker has rated them yet.
05

Roadmap

Planned, not built yet. No dates.

Planned

Share to Sloth-Butler

Send a message straight from email or a chat app instead of copy-pasting.

Planned

Reminders on your phone

A notification when a follow-up is due, instead of checking the board.

Planned

Export your log

Download your protected hours and money as CSV for your own records.

Planned

Shared terms for a team

One set of rates and floors that everyone on a small team replies from.

Planned

More languages

Italian, Dutch, Hindi and others, with native-speaker review.