General usage
Learn how to use the Truffle client interface effectively
Selecting an App
Choose which sandbox of tools (app) your agent will use to complete the given task.

App Selector
Use the app selector to slot your model into the appropriate toolkit before submitting your prompt.
- Your prompt will be classified as you type, assigning a relevant app based on its contents.
- Use option/alt to open the app select menu and manually select an app.
You can start the process of creating a new app by selecting the “Create app” option. See our SDK docs for more info.
Writing and Submitting Prompt
Compose your instructions to the agent in the prompt bar. You can attach external files to include in the agent’s context.

Prompt Bar
The prompt bar lets you type, attach files, and view an autogenerated summary title for long prompts.
Long prompts will have summary titles generated. These are not used by the agent; they exist for the user’s experience.
Running Tasks
A task is an ongoing loop of reasoning stages, tool calls, and result outputs. Your agent iterates through steps until it’s ready to respond.

Task View
The task view displays each reasoning step, tool call, and result so you can track your agent’s process.
A step is an individual reasoning stage, tool call, and result your agent generates within a task Your agent will iterate through steps until it’s ready to provide results to the user via the Respond To User tool
Pause and Lock
Control the task execution flow by using pause and lock features directly in the task view.

Pause & Lock Hotkeys
Press P to pause a task (halts new step generation), or L to lock a task (forces continuous generation until unlocked).
Use pausing when you need to inspect current progress; use locking when you want uninterrupted deep analysis.
Viewing Results
Each step shows a result based on the tool used. When the agent uses the Respond To User tool, that result is delivered back to you.

Results Preview
Click on a step in the task view to open a preview window showing the reasoning, tool calls, and results.