Build an applet
An applet is a small web app that an agent writes for you, such as a tracker, a calculator or a dashboard. It lives in Applets on the rail. It can store its own data and use your connected apps. You don’t need to touch the code, but you can read it.
Build one
- Go to Applets > New applet. A new task opens with “Make a new applet that ” already typed. In a workspace home you can also choose Build > App or Build > Dashboard.
- Finish the sentence. Say who uses it, what they do in it, and what it shows. For example:
Make a new applet that lets the Harbor & Pine team log property viewings. Each viewing has an address, date, agent name, client name and a rating from 1 to 5. Show a table of this week’s viewings, sortable by date, and a count of viewings per agent.
- Send it. The agent may ask about details, or ask to use a connector. Answer in the thread.
- When it’s done, open the applet from the task or from Applets, and click Open to run it.
Describe the screen, not the technology. “A table of this week’s viewings” is enough. Start with one screen and add features one at a time in the same task. A ten-screen app built in one go is hard to check.
Open it beside the list
In Applets, your applet is listed with the others in your organization, such as the built-in Templates applet. Use the tabs All, Your applets and Shared with you, or search. The list may start on your current workspace; to see others, pick another workspace, or All workspaces, from the workspace menu in the list’s header.
Click an applet to run it in a pane beside the list. On a phone it opens full screen. It’s a working tool: what you change in it is saved, as long as the applet stores its data, which the agent sets up when the applet needs it.
When several people use it at once
An applet’s shared data is the same for everyone who can open it, and each saved change records when it was made and by whom. An applet doesn’t refresh on its own when someone else changes something, unless it’s built to. If several people use it at the same time, ask the agent to refresh the data every few seconds and show who changed it last, for example “Updated by Priya Raman · just now”.
Change it and keep versions
On the applet’s Overview, click Edit in a task and describe the change, such as “Add a filter by agent name”. Edit in a task is shown to people who manage the applet.
Every publish creates a new version. The first is “Initial version”, and each later one carries the agent’s summary of the change. Click the version badge in the applet’s header, such as v2, to open Versions: “Every published build of this applet. Select one to see it running; restoring republishes it as a new version.” You can also open Versions from the … menu.
If a change breaks something, select an earlier version and click Restore. The old build is published again as a new version, and the current version stays in history, so nothing is deleted. Only people who manage the applet can restore.
Share it
Before you share an applet, enter a few rows of real data yourself to check it works. Then click Share in the applet’s header, or open its Access page and click Share there. Add people, agents or user groups, and under General access choose who in the workspace or organization can open it. Viewer lets them open and run the applet; for example, everyone in Northwind Advisory can open the tracker as a viewer.
The roles you can pick depend on your organization’s permissions system. With the newer permissions system they are Viewer and Editor, and General access reads “Everyone in [workspace] can view”. On other organizations they are Viewer and Admin: Admin can also configure it, manage access and delete it. The creator is the Owner.
The Overview shows whether it is personal, published to the workspace or published to the organization. To publish to the whole organization you need to manage the applet, and it must already be shared with a workspace. With the newer permissions system, share it with your organization under General access instead.
Security checks on every publish
Every publish runs automatic security checks. People who manage the applet see the results on its Security page: “Checks run at every publish. Findings stay until a re-scan confirms the fix.”
- Checks shows how many checks passed, such as “5 of 5 passing”. There are up to 10 checks; the ones that don’t apply to the applet, such as backend or connector checks for an applet without them, are skipped and not counted.
- Reach shows what the applet can reach. For an applet with no connectors, no external hosts and no workspace AI it reads “Nothing reaches past the applet sandbox.”
- Run checks runs them again. Run deep scan reads the source and runs nothing, and takes about 2 minutes.
By default the checks don’t block a publish: they run after it’s saved, and a failed check never undoes it. They’re there so you can see and fix problems.
The applet’s pages
- Overview, Activity and Versions: what it is, what its runs cost, and every published version.
- Evaluations: the tests that grade it, where monitoring is on.
- Code: the code the agent wrote.
- Dispatched tasks: tasks it started for you.
- Access: who can use it.
- Tables and Storage: the data it keeps (for people who manage the applet).
- Resources, Connectors and Secrets: the files, apps and keys it uses.
- Security: scan results (for people who manage the applet).
- Settings.