Neurons

A neuron is the building block of a workflow - it takes something in, has the AI work on it, and sends the result on.

A neuron is one step of a workflow. It takes something in, has the AI work on it with your instructions, produces a result and sends that result on. Use one neuron for one clear job, and connect several neurons when a task has more than one step.

The five stages of a neuron

Every neuron has the same five stages, shown left to right on its card:

START: when the neuron runs. Press Play, set a schedule, start it from another neuron, from your own systems, or from an app event.
GET: what it takes in. Text, files, recordings or a table.
DO: what the AI does. Your instructions, the apps it may use, and what it reads from Memory.
MAKE: what it produces. Text, an image, audio, a table or a decision.
SEND: where the result goes. The next neuron, email, Telegram, your own systems or a table.

See Neuron parts for each stage's settings.

Add a neuron

There are three ways to add a neuron:

From the top bar: click Add, then Neuron. The neuron appears on the canvas, named "Neuron" plus a short code, with its DO settings open and one required text input ready.
From the canvas: double-click an empty spot to add a neuron there, or click + (Add a neuron after this one) on a neuron to add the next one, already connected.
From the Neurons section: click Add neuron to answer five short questions first: When should it run?, What does it take in?, What should the AI do?, What does it produce? and Where does the result go? Use Next and Back to move between them, then Create neuron. Skip, add blank adds an empty neuron instead.

The neuron card

The neuron card shows each stage with its current setting, so you can read a neuron at a glance:

the neuron's name, and how many runs it has had and what a run costs on average;
each stage's summary, such as the trigger in START, the inputs in GET, the instructions in DO, the result type in MAKE and the destinations in SEND;
the last result under MAKE, or NO RUNS YET before the first run.

You can change some settings right on the card: click the DO instructions (Click to edit) to edit them, type a value into a text input, upload a file, record audio or pick a table.

Collapsed neuron cards

A collapsed neuron card folds the neuron to a short summary of each stage and its last result. Use Collapse neuron / Expand neuron on the card, Collapse all neurons / Expand all neurons in the top bar's More menu, or turn on Auto-collapse in the view controls.

The neuron settings panel

Click any stage on a neuron card to open the neuron's settings panel at that stage. The panel has a tab for each stage: Start, Get, Do, Make and Send.

Change the neuron's name in Neuron name at the top of the panel.
Delete the neuron from Neuron options › Delete neuron.
Click Close, or press Esc, to close the panel.
Make the panel wider: drag its left edge, or click Expand panel at the top for a wide panel and Restore panel size to go back. With the left edge selected, the arrow keys resize it, and Home and End make it as narrow or as wide as it goes. Double-click the edge to go back to the usual width. When the panel is already about as wide as it goes, Restore panel size takes it back to the usual width. When a wide panel leaves too little room beside it, the canvas controls and the minimap move to the bottom-right corner, below the panel. The panel keeps its width in this browser. On a narrow window or a phone it keeps its usual size.

The edges of a neuron card

The edges of a neuron card are where connections start and end:

Receptor: the left edge. Other neurons connect into it.
Transmitter: the right edge. Click it, then click another neuron's receptor, to connect the two. A decision has one transmitter per condition.
On error: the red dot on the right edge. Click it, then click another neuron's receptor, to run that neuron if this one fails. See On error links.

Each neuron in the Neurons section shows its status (Ready, or how many issues it has to fix) and a menu, Run and options, with Play, Test this neuron only, Play all queued (when an input has a queue), Configure (opens the settings panel), Collapse on canvas, Rename, Ask Cortex about this (see Cortex), Memory settings and Delete neuron. Each flow has its own play button (Run the flow).

Issues on a neuron

When a neuron has a problem that would stop it running, such as missing instructions or a broken reference, the neuron shows the issue on its card, as a dot on the settings panel tab where it lives, and as "1 to fix" (or more) in the Neurons section. The tab shows the issue at the top, for example "DO needs instructions for these inputs". Some issues have a Fix button that takes you to the setting, or Open Memory when the problem is in Memory.

Common problems

A neuron shows "to fix" in the Neurons section

A neuron that shows "1 to fix" (or more) in the Neurons section has an issue to fix before it can run. Choose Configure in its menu, open the tab with a dot, and read the highlighted issue at the top.

Neurons can't be placed on top of each other

Neurons can't overlap on the canvas. Drag the neuron to an empty spot.

A neuron I deleted is still needed by another neuron

When you delete a neuron, connections to it go too, and any links that passed its values become broken. Open the neurons it was connected to and fix the broken links. See Connections.