AI agents

AI agents that are actually yours.

An agent is just a small app that works on its own — on a schedule, or the moment something happens — using your own AI to think, and pinging you when it matters. You don't write any code: you ask your assistant in plain words, and it builds the agent on your private server. Your server, your AI, your data — one flat price, no per-agent fees.

7-day free trialThen $29/monthRuns on your own Claude or OpenAI account

How agents work

Software that does the job before you ask.

An agent is an app that works on its own — nobody has to open it, click it, or remember it. Give it a time to run or an event to watch, let it think with your own AI, and the repetitive work quietly takes care of itself. And you set it all up by asking your assistant, not by writing code.

It runs on time, on its own

Tell it when — every morning, every hour, every Monday — and the job just happens. Nothing to remember, nothing to start by hand, and it keeps running even after an update or a reboot.

It reacts the moment things happen

A form comes in, a payment lands, another tool pings it — and your agent springs into action right then. No waiting, no checking, no one refreshing a page to see if there's work to do.

It can read, write, and decide

Plug in your own Claude or OpenAI account and your agent handles the messy, wordy work — reading, summarising, drafting, judging — not just shuffling data around. Your subscription, never a resold one.

Something sets it off a set time · or an event
Your agent does the work checks things · asks your AI
It acts for you emails · alerts · updates an app

At a set time, or when something happens, your agent does the work — using your own AI when it needs to think — then emails you, alerts you, or updates an app.

Examples

Small helpers that do one job well.

Morning briefing

Each morning it gathers what you care about — today's bookings, new form responses, your calendar — has your AI boil it down, and sends you a one-page brief before you're at your desk.

Lead qualifier

When someone fills in your form, the agent scores the lead with your AI and drafts a first reply, then flags the ones worth calling first.

Inbox triage

A new message arrives; the agent sorts it, drafts a reply in your voice, and leaves it on a private page for you to approve.

Weekly report

Every Friday it pulls the week's numbers from your other apps, writes the summary, and emails it to whoever needs it.

Watchdog

It checks a price, a stock level, or a web page on a timer and pings you the moment something crosses a line you set.

Content drafter

Feed it your notes through the week; every Monday it drafts posts or a newsletter for you to edit and send.

Example prompts

Describe the agent. Your assistant builds it.

Paste one of these into an assistant connected to your croft, and tweak the timing and the details to fit. Your assistant writes the app, wires up the schedule, and deploys it.

Daily briefing agent

Build me a daily briefing agent. Each weekday at 7am, collect the day's bookings and any new form submissions from my other apps, using my own OpenAI account (I'll add the key in settings) to summarise them into a short plain-English brief, and email it to me. Keep the last month of briefings on a page behind my login.

Reminder emails

Build me an app that sends reminder emails on a schedule. Let me create reminders with a message, a recipient, and a rule — a specific date, or every week or month. Each morning, check what's due, send those emails, and mark them sent so nothing goes twice. Give me a simple page to add and edit reminders.

Why Croft

An agent needs a home. This is a good one.

Your own AI

Connect the Claude or OpenAI account you already use. We don't resell AI or mark up usage — the thinking is billed by your provider, not by us.

Runs on your server

Your agent lives on a private machine that's yours alone — not a shared, multi-tenant pool you share with strangers.

It can remember

Every agent has its own database, so it keeps state between runs — what it's seen, what it's done, what to do next.

It can have a face

Because an agent is just an app, it can also have a private dashboard behind your login — not only a job that runs in the dark.

One flat price

No per-agent fees, no per-run metering. The flat price covers the ground; your AI account covers the thinking.

Yours to take

Export the whole thing — code, database, schedules — and run it anywhere Docker does. No lock-in, ever.

One honest note

Agents that think use your own AI.

To read, summarise, and decide, an agent uses your own Claude or OpenAI account — the same kind you may already pay for. You connect it once (your assistant walks you through it), Croft keeps it safe, and the AI usage is billed by Claude or OpenAI, not by us. Same bring-your-own-AI deal as the rest of Croft: no markup, no lock-in.

And plenty of useful automations need no AI at all — reminders, overnight syncs, scheduled reports. Those run on the flat price alone.

Stake out your croft.

Your team's first app could be live before lunch.

Get your croft

7 days free, no card to start. Then $29/month — cancel anytime and take everything with you.