> ## Documentation Index > Fetch the complete documentation index at: https://docs.moneypenny.ai/llms.txt > Use this file to discover all available pages before exploring further. # Journeys > Design the paths a call can take — the jobs your agent handles and the steps inside each one. A journey is a job your agent can handle on a call: taking a message, booking an appointment, answering questions, or putting the caller through to a person. Steps are the stages inside a journey — collect the caller's name, then their order, then send the confirmation. The **Journeys** tab shows all of this as one diagram, **Your agent's journeys**, with each step as a card you can move, link and edit. ## How your agent follows the diagram Every agent has a default step called **General conversation** — the catch-all where the call starts and where your agent comes back when nothing more specific is going on. When a caller says what they want, the matching journey begins: a journey's first step carries a plain-English condition that tells your agent when to start it ("The caller wants to place a bread order"), and steps that follow another step start automatically once the one before them is done. When a journey finishes, your agent either returns to the general conversation or ends the call politely — the small caps at the bottom of each journey show which. The [build chat](/building/build-with-ai) creates journeys for you from a description; this canvas is where you see exactly what it built and fine-tune it by hand. ## Working the canvas * **Add step** creates a new step and opens it in the settings panel, ready to name. * **Click +** under a step to add the one that follows it. * **Drag a step onto another** to move it underneath. * **Drag from a dot** on one step to another to link them. * **Select a step and press Delete** to remove it — any steps that followed it are unlinked, not deleted. The General conversation step can't be removed. * **Tidy up** rearranges everything into a neat layout after you've been dragging things around. The controls in the bottom corner zoom and fit the diagram. Steps can branch — two steps can follow the same one, each with its own condition — but they can never loop back into each other. ## Step settings Pick any step and the **Step settings** panel opens beside the canvas: * **Name** — what this step is for, in a few words. * **When does this step start?** — shown for steps that begin a journey: the plain-English cue your agent listens for. Steps that follow another step don't need one. * **Instructions** — what your agent should do during the step. The step completes once its instructions are done and every detail marked "must be collected" has been captured — and your agent never re-asks for anything it already has. * **If something your agent must do fails** — carry on with the call, or end the call politely. For example, if your agent can't send a booking to your calendar. * **Information to collect** — the details your agent gathers in this step. Each one has a name, an optional note on how to ask for it ("a UK postcode"), and a **Must be collected** tick if the step shouldn't complete without it. * **Actions** — things your agent actually does in this step: sending an email, creating a booking, transferring the call. Actions that work through your [connected apps](/apps/connect-apps) can be added before the app is connected and wired up later. * **Remove step** — takes the step out of the journey. Like everything in the editor, canvas changes save to your draft — try them from the [Test tab](/building/testing), then [publish](/building/publishing) when a real call should follow the new journey.