When to Use It

• Ask the Main Agent to read project files and draft a research plan, task list, or deliverable structure.
• Use it when a task spans literature, data, analysis, writing, or knowledge graph work and needs subagent coordination.
• Use it to turn existing material into executable steps, then review progress over multiple rounds.

Steps

1. Click Open Project in the toolbar and choose the research-project folder. For an existing project, use the same entry to Switch Project. The software prepares the .agents folder and the default DuckDB.
2. Click Select Specialist Agents below the Main Agent title and enable only the specialists needed for this task. Enabled means callable by the Main Agent; it does not switch the active workspace.
3. To manage or add capabilities, open Settings → Skills and review names and purposes by Agent. Built-in Skills sync when the project opens, so basic capabilities do not require manual import.
4. Enter a task prompt at the bottom with the goal, material locations, output format, permission boundary, and acceptance criteria. Before sending, you can switch among enabled models and permission modes in the input area.
5. Click Send. During execution, collapse the process, inspect tool results, and search the current conversation. Afterwards, start a new conversation, restore history, or clear the conversation.

Acceptance Checks

The Main Agent response should state what was completed, which project materials were read, which specialist Agents were called, where results were saved, and what still needs confirmation. For database, file-writing, or web tasks, also review the execution process, sources, and permission mode. Open the file workspace on the right to inspect code, SQL, and previews.