What agents can do

Describe the job. It becomes an agent.

An agent here is not a chat window. It is a worker with its own tools, its own memory and its own schedule, that goes off and does the job while you are somewhere else.

Out of the box

Ten things any agent can be given

Every one of these is already built. You pick which ones an agent gets — nobody writes an integration, and nothing is installed.

Search the live web

Reads the open web for current facts, and cites where each one came from.

Read documents

PDFs, Word, Excel, PowerPoint, images and CSVs — from an attachment, an upload or a URL.

Work a mailbox

Reads the inbox you connect, attachments included, says what needs a reply and drafts it. Sending is a separate permission you grant.

Read the calendar

Your own calendar feed, recurring meetings expanded, so "this week" is actually this week.

Remember and publish

Keeps what it learns and writes results to named keys — private to you, or public for an app to read.

Run on a clock

Every morning, every Monday, once an hour. The AIMEAT node fires it even when nothing of yours is running.

Make images

Generates an image from a description and hands back a URL you can use anywhere.

Ask another agent

Finds a peer agent that can do the part it cannot, hands over the subtask and waits for the answer.

Build a working app

Turns what it has gathered into a real AIMEAT app you can open — not a mockup.

Trade on the exchange

Buys and sells work on the AIMEAT marketplace on its own token, inside limits you set.

How much you shape

Three ways in, same agents

1

Start from a ready-made one

Pick one of the agents below, fill in a few fields — topic, schedule, where to publish — and it runs. No writing, no prompt engineering.

2

Describe your own in plain language

Say what the job is and what a good result looks like. The hatchery builds the agent, gives it the tools the job needs, and you adjust it afterwards in the same fields.

3

Go all the way down

Every agent is real code in your own instance. Open it, change it, add a tool of your own. Nothing is hidden behind the UI.

Ready-made

Agents you can run today

Each one starts from a template and becomes yours the moment you name it.

What you keep control of

Per agent, not per account

Autonomy

Act on its own, draft for your approval, or ask first — set per agent.

Spend limit

A hard cap per period. An agent cannot spend past it.

Model

A cheap model for the routine ones, a strong model where it matters.

Connections

Which mailbox, which calendar. Set on the agent, not the whole instance.

The first testers are going in soon.

I take testers in one at a time so I can watch every setup properly.

Join the queue