Glossary
This section explains terms that appear frequently throughout this manual and in the user interface.
Creating
Project
The workspace for building one application. When you open a project, its own container is launched, and the conversation history and source code are kept together. The content is preserved unless you delete the project.
App
The deployed result produced from a project. This is what your colleagues open via a link.
A project is the place for building, while the app is the built result. Changes made in the project do not affect the app until you click Deploy.
Template
The base form from which a new project starts. A few built-in templates are available, and you can also save a screen you've created as your own template to reuse later.
Catalog
A list screen showing all deployed apps. You can set sharing targets here → App Catalog and Sharing
Ways to Interact
Builder (Build)
A mode for creating and modifying apps through chat. The agent writes directly to the project's source code.
Ask
A mode for querying data only, without building an app. The platform is queried in read-only mode and answers are provided in tables and charts → Chat-based Data Query
Plan
A step where you first decide what to create and how before building a large screen. Listing the screens, widgets, and required data for review and editing before instructing the build helps reduce back-and-forth.
Review Mode
A mode that does not automatically finalize the agent's edits but instead lets a person choose to accept or revert them → Developing Apps with the Builder
What Runs Under the Hood
Agent
A component that receives a user's request, selects and invokes tools, observes the results, and then chooses the next tool—repeatedly. It uses file reading/writing, command execution, and platform queries.
Coding Engine
The part that actually writes code. It is pluggable—you can select between the built-in engine and an external coding agent in the environment settings.
Self-Verification
The procedure where an agent checks its own work after modifying files. It runs type checks, reads failures and re-fixes them, and then actually launches the screen. This back-and-forth can happen multiple times within a single request, making the first response slow.
Session Container
The container dedicated to a project that is launched when you open the project. The development server runs inside it, and the agent's commands are also executed only within it.
| Default | |
|---|---|
| Memory | 2 GB |
| Idle Stop | 30 minutes |
🔴 Idle Stop is NOT "deletion." Only the container shuts down; the source code and conversation history remain. When you reopen the container, it starts again—which is why projects that haven't been opened for a while may take longer to load initially.
Docker Proxy
The Docker interface used to create and delete session containers. Instead of exposing the server's Docker socket directly, operations go through this proxy—because exposing it directly would provide full host-level privileges.
Viewing and Exporting
Preview
The screen that builders see. Since it shows the development server directly, changes made by the agent are reflected immediately. The address is /preview/….
Deployed App
The built result that appears after clicking Deploy. This is what colleagues see, and it only changes when Deploy is clicked again. The address is under the same domain as the Studio, at /container/….
Version / Revert
Each deployment creates a new version (10 versions are kept by default). If a new deployment is faulty, you can revert to a previous version.
Access Scope
Private apps can only be opened by owner, admin, and users who have been granted access. Knowing the link is not sufficient to open one.
Automated Processes
Watcher
Automation that has no user interface. You can write monitoring conditions in natural language—like "notify me if the pressure on Line 3 exceeds 90"—and it checks and notifies you periodically → Watchers and Notifications
Field Skill
A feature that lets you document an engineer's expertise and save it. The agent refers to it when relevant. It persists even after the person leaves and can be handed down to other sites → Field Skills
Accounts and Permissions
Roles
| Role | Responsibilities |
|---|---|
| admin | Environment settings · User management · AI providers |
| builder | Builds and deploys apps (default on account creation) |
| viewer | Views deployed apps |
🔴 The default role is builder. Please assign viewer to users who only need to view → Work Process
MCP (Model Context Protocol)
An open standard that allows AI agents to discover and invoke external system functions as "tools." The Studio uses this standard to read platform data.
Terms from Other Products
PlantPulse shares the asset hierarchy, tags, and data points across its four products. Terms not listed here are in the other three manuals—Platform Glossary · Edge Glossary · AI Glossary.