Skip to content
mintbot.ai

mintbot docs

Tutorials, guides, and technical documentation for mintbot.ai — the product itself lives there; this site is just the docs.

Topics

  • Getting started — pick a plan, pay, and your agent deploys itself in minutes. No setup required — a screenshot-led walkthrough from checkout to your first chat.
  • Best use cases — when to use which LLM provider, and how to plug in a ChatGPT or Claude subscription for the cheapest LLM-per-dollar.
  • Projects — give each piece of work its own folder on the agent's server, with its own files, instructions and conversations.
  • Share your agent — invite people to use your agent: everyone gets their own chats, language and Telegram, on one agent and one credit.
  • Several agents at once — four ways to use more than one agent side by side in Telegram, from switching in the main bot to a bot of your own per agent.
  • Name your agent in Telegram — show your agent under the name you want in your Telegram chat list, using a simple group trick.
  • Your own Telegram bot — connect a bot you create with @BotFather, so your agent answers under your own name, @username and picture.
  • Moving data between agents — copy memory, skills, personality and chat archive from one of your agents into another.
  • How mintbot handles files — why every upload lives on your own server forever, and how the LLM gets a format-shifted copy on demand.
  • MintOffice — the Brand Partner portal and its API.
  • Developer API — a REST API for ordering and managing agents from your own code: API keys, a EUR wallet, signed webhooks.
  • MCP server — the same API as tools for Claude Code, Cursor and other MCP clients.
  • BeeOffice ↗ — referral programme; earn 20 % of mintbot's lifetime profit from every client who signs up through your link.
  • Meerkat ↗ — public status page showing the live health of every mintbot service, plus the last 7 days of history.

Frequently asked questions

"Why would I want a mintbot agent if I already pay for ChatGPT or Claude?"

Short answer: mintbot doesn't replace your favourite LLM — it gives that LLM hands.

ChatGPT and Claude chat are brilliant brains in a tiny box. Ask anything, get an answer back, conversation ends. Files dropped into the chat eventually age out. The model can't open a terminal, schedule a cron job, or remember last quarter's research without you re-uploading it.

A mintbot agent fixes that. Under the hood it talks to the same models you already know — GPT-5 via your ChatGPT subscription, Claude Opus via your Anthropic subscription, or whichever model you point its BYOK at. But around the LLM, mintbot wraps a toolbelt:

  • A personal VPS the model can actually drive — read/write files, run code, install packages, schedule cron jobs, keep long-lived browser sessions alive.
  • Telegram-first messaging, so your agent is just another contact in your phone — available 24/7 from the same app you already live in.
  • Persistent file archive — every photo, document, voice note or paste you send gets stored on your VPS forever, and an LLM-optimized version is handed to the model on demand. (How mintbot handles files explains why this is a bigger deal than it sounds.)
  • Skills — composable mini-programs the agent learns over time, from "log this expense to my Google Sheet" to "scan my Polymarket positions every morning".
  • More than one channel — the same agent answers in Telegram and in its web panel, and on request it sets itself up on other messaging apps too.

The combinations are where it gets interesting. ChatGPT can answer a question about a PDF you uploaded — once. A mintbot agent has that PDF forever, can re-process it next month with a different model, extract a table out of it and email the table to your accountant, then compare it to last quarter's version. Frontier-model intelligence × persistent tools = capabilities neither has alone.

Think of mintbot less as "an LLM chat product" and more as "your personal computer that happens to be staffed by a brilliant intern" — and the intern is whichever frontier model you point them at.

"What's the cheapest LLM credit I can buy?"

Short answer: you already have it. If you pay for a ChatGPT or Anthropic subscription, your mintbot agent can run on that quota through the bundled Codex or Claude Code CLI — no per-token billing, no top-ups.

The default mintbot proxy is convenient (works out of the box, no API key) but bills retail per token, which makes it the most expensive option long-term. ChatGPT Pro at $200 / month buys dramatically more model usage than $200 of raw API tokens — the break-even is usually well under 10 days of normal use, and every call after that is free at the margin. Anthropic Pro / Max works the same way through Claude Code.

The wiring takes about a minute from the web panel: open LLM providers, click Add LLM provider, pick ChatGPT subscription or Claude subscription, press Sign in and finish the login in a browser tab. Done.

Full breakdown and a step-by-step in Best use cases.

Found something wrong or missing?

These docs are written for the people who actually use the product, so gaps and stale spots get fixed fastest when you flag them. Just tell your mintbot agent what's confusing, wrong, or missing — they'll send the feedback our way and we'll update the page.