---
title: A worker’s plan
description: Read and edit a Worker's plan — the chart of steps Ysra drafts (When, Read, Do, Decide, Act, Deliver) and each run lights up.
new_in: "3.0"
---

A worker’s **plan** is a chart of the steps it goes through: what starts it,
what it reads, what it does and decides, what it acts on, and what comes out.
Ysra drafts the plan; you change it and save it. On every run, the worker goes
through the steps in order and decides how to do each one itself. At a
**Decide** step it answers the question and follows the matching path.

The chart is the plan, not a program. It gives the worker no new permission:
what a worker may read and do is set when it is started (see
[Start a worker from Ask](create.md)). A step can ask for more approval, but
never for less.

## Where the plan is

The plan is on the worker’s page, in the web app and the desktop app: open
**Workers** and choose the worker. The chart is under **Plan**, between what
the worker may do and its runs. It is saved on its own, with **Save plan**;
**Edit** changes the rest of the worker (see
[Edit a worker](create.md#edit-a-worker)).

Beside **Save plan**, the plan says which version it is (for example
"Version 3"), or "Unsaved changes" while you edit.

## The six kinds of step

| Step | What it is for |
|---|---|
| **When** | What starts it. A plan has exactly one, at the start. |
| **Read** | A file, a link, a connection or a mention. |
| **Do** | An instruction in plain words. |
| **Decide** | A yes or no question, and two paths. |
| **Act** | Send, post or change something. |
| **Deliver** | What comes out, and where it goes. |

The steps are laid out left to right, starting from **When**.

## How a plan is drafted

Ysra drafts a plan:

- **For a worker started from Ask.** A worker started from its card in a
  conversation starts without a plan. Ysra drafts one from the worker’s own
  instructions when you ask.
- **For a worker made before plans existed.** It has none either, and Ysra
  drafts one the same way.
- **On the step-by-step screen.** A worker described on the older
  **New worker** screen gets a drafted plan on **Plan it**. See
  [The step-by-step screen](create.md#the-step-by-step-screen).

For a worker with no plan, Ysra drafts only when you ask. The page says "This
Worker has no plan yet. A plan shows the steps it goes through, as a chart you
can change." Press **Draft a plan**. If no plan can be drafted, the page says
why and the button becomes **Try again**.

The draft is shown with "Drafted from what this Worker does. Change anything,
then save it to make it the Worker’s plan." and marked "A draft, not saved
yet". Nothing is saved until you press **Save plan**. To drop the draft, press
**Discard changes**; the page goes back to **Draft a plan**.

If Ysra can’t draft a plan, the worker gets a plain three-step one: **When**,
then **Do the job** (its instructions), then **Report what was done**.

## Edit the plan

<div class="steps" markdown>

### Add a step

Press **Add a step** and choose a kind. If a step is selected, the new one is
placed after it and linked from it when that step has no path out yet.
A plan has only one **When** step, so **When** isn’t offered once there is one.

### Link the steps

Drag a step’s right edge to another step to connect them. A step has one path
out, so a new link from it replaces the old one. Only a **Decide** step
splits: it has two outputs, **Yes** and **No**, and each leads to its own
step. Nothing can lead into **When**.

### Set up a step

Click a step to open its settings:

| Step | Settings |
|---|---|
| Every step | **Title** (for **Decide**: **The question**), up to 120 characters |
| **When** | **Runs**: **Only when run by hand**, **Weekdays at 09:00**, **Every day at 09:00** or **Every hour** |
| **Read** | **From**: **A file**, **A link**, **A connection** or **A mention**, then the **File**, **Link**, **Connection** or **Mention** |
| **Do** | **What to do** |
| **Decide** | **How to tell** |
| **Act** | **Action** (**Send an email**, **Draft an email** or **Post a message**) and **Before it acts**: **Asks first** or **On its own** |
| **Deliver** | **To**: who or where |

Other steps have an **In plain words** box. Press **Done** to close the
settings.

Changing **Runs** on the **When** step changes when the worker runs, once you
save. **Runs** offers only these four choices; any other schedule comes from
the words you started the worker with, or is set under **Schedule** on the web
app’s Edit screen. See [Schedules](schedules.md).

### Remove a step or a link

Select a step or a line and press Delete. In a step’s settings,
**Remove this step** does the same. Removing a step removes its links too.

### Save

While you edit, the chart says "Unsaved changes" and offers
**Discard changes**. Press **Save plan**. Every save makes a new version of
the worker ("Plan saved as version 4"), and runs already started keep the
plan they started with.

If the worker was changed somewhere else in the meantime, the save is
refused instead of overwriting that change, and the plan says so. Make the change again on the latest
version.

</div>

You can drag steps around to tidy the chart; their places are saved with it.
An archived worker’s plan can’t be changed. Its page still shows the chart,
read-only.

## Asks first

An **Act** step is set to **Asks first** unless you choose **On its own**.
**Asks first** makes that action stop for your approval, even if the worker’s
grants would let it run on its own.

It only ever tightens. **On its own** doesn’t loosen anything: sending mail
and posting to Slack always ask you first, whatever the chart says. See
[Runs and approvals](runs.md).

## Steps that can’t run yet

A step that needs something the worker hasn’t been given is marked on the
chart with why, and the chart says so above it: "One step can’t run yet" (or
"3 steps can’t run yet"). For example:

| The step | What the chart says |
|---|---|
| **Read** from a link | "Reads the web: this Worker needs web access." |
| **Read** from a connection | "Reads a connected source: add the connection to this Worker." |
| **Act** that drafts email | "Drafts email: this Worker needs to be allowed to draft mail." |
| **Act** that sends email | "Sends email: this Worker needs to be allowed to send mail." |
| **Act** that posts to Slack | "Posts to Slack: this Worker needs to be allowed to post." |
| **Act** that posts a message | "Posts messages: this Worker needs to be allowed to post." |

A run skips such a step and says why. Change the step or remove it; the mark
goes once every step needs only what the worker has. If the worker really
needs more, ask for it again in Ask with a job that says so, and start the new
card. See [Start a worker from Ask](create.md).

## A plan that can’t be followed

Before you save, the chart lists below it what would stop the plan from being
saved, such as "The steps go round in a circle." or "“Read a source” can’t be
reached from “When”."

If the plan still can’t be followed, saving is refused with one plain
sentence, shown above the chart as it is. The sentences are:

- "A chart starts with exactly one “When” step."
- "Nothing can lead into the “When” step."
- "A step can't link to itself."
- "Two links join the same steps."
- "A “Decide” step has one or two paths out."
- "Both paths out of a “Decide” step need a label."
- "Only a “Decide” step can split into two paths."
- "The chart goes round in a circle."
- "Not connected to the start: …", naming the steps that can’t be reached.

Change the chart and save again.

## Follow a run on the chart

Once a worker has run, **Plan** has two views: **The plan** and
**Latest run** (**This run** while it is still going). The worker’s page
opens on the run while it is still going, and on **The plan** otherwise.

The run view shows the saved chart, read-only, with the run’s state above it
(for example **Working** or **Done**). It lights the chart up step by step,
and keeps up as the run goes on. Select a step to see
its kind and how far the run got with it. The legend under the chart says what
each colour means:

| Label | Meaning |
|---|---|
| **Done** | The worker finished this step. |
| **Working now** | The worker is on this step. |
| **Waiting for approval** | The step stops for your approval. See [Runs and approvals](runs.md). |
| **Skipped** | The worker passed over it, for example the other path of a **Decide** step. |
| **Not reached** | The run didn’t get to this step. |

The lines the run took are highlighted. If a run didn’t mark its steps, the
view says "this run didn’t report its steps".

On the phone, a worker’s page shows the chart read-only, with what
the latest run did at each step.

## Related

- [How Workers work](how-workers-work.md) — what a run does with its steps.
- [Runs and approvals](runs.md) — following a run and approving a step.
- [Start a Worker from Ask](create.md) — where a plan comes from.