Cowork Workspace
The primary workspace for sending tasks, monitoring agent execution, and reviewing results in the AssistMe Desktop App.
Overview
The Cowork tab is the heart of the Desktop App — it's where you interact with the AI agent to get real work done. Unlike simple chat, Cowork provides a live task execution environment where you can watch the agent work in real-time, including tool calls, browser screenshots, code execution, and more.
Sending a Task
- Type your prompt in the text area at the bottom of the screen
- Press Enter or click the Send button to submit
- The agent begins working immediately — you'll see live updates in the workspace
Tips for effective prompts:
- Be specific about what you want the agent to do
- Mention file paths or URLs if relevant
- Describe the expected outcome
Stopping a Running Task
While the agent is processing, the Send button turns into a red Stop button. Click it to cancel the current task. The message will be marked as "Cancelled" in the conversation history.
Live Execution View
The Cowork workspace uses a split-panel layout:
| Panel | Content |
|---|---|
| Left (Messages) | Conversation messages — your prompts and agent responses |
| Right (Logs) | Real-time event stream showing exactly what the agent is doing |
You can drag the divider between panels to resize them (the log panel width ranges from 240px to 700px).
Message Panel (Left)
Shows the conversation flow:
- Your messages — The prompts you've sent
- Agent responses — The final output from the agent
- Status indicators — Pending, Running, Completed, Failed, or Cancelled
- Auto-scroll — The panel automatically scrolls to the latest message
Log Panel (Right)
The log panel is a desktop-exclusive feature that provides real-time visibility into agent activity:
- Tool Calls — Each tool invocation is shown with its name and expandable parameters
- Tool Results — Output from each tool call, with file size indicators for large results
- Browser Screenshots — Full-size screenshots captured during browser automation tasks
- Thinking Blocks — Expandable sections showing the agent's reasoning process
- Timestamps — When each event occurred
Click any event to expand its details. This is invaluable for understanding how the agent accomplished a task, not just the final result.
Adaptive Polling
The Desktop App uses smart polling to balance responsiveness with efficiency:
| State | Poll Interval | When |
|---|---|---|
| Active | 1 second | A task is currently being executed |
| Recent | 2 seconds | Activity detected in the last few seconds |
| Idle | 4 seconds | No recent activity |
This reduces server load by up to 80% during idle periods while keeping real-time responsiveness during active task execution.
Browser Screenshots
When the agent performs browser automation (navigating pages, clicking elements, filling forms), screenshots are captured and displayed inline:
- Screenshots appear in both the message panel (as part of the response) and the log panel (as events)
- Full-size rendering lets you see exactly what the agent sees
- Useful for verifying that the agent is interacting with the correct elements
Session Management
Starting a New Session
Click the "+ New" button in the header or press Cmd/Ctrl + N to start a fresh session. The previous session is preserved in your history.
Browsing History
Click "History" in the header to open the session history panel. Browse past sessions and click any to resume it.
Sidebar Session List
The sidebar lists your recent sessions with quick actions:
- Search — Filter by keyword
- Star — Pin important sessions
- Rename — Give a session a descriptive title
- Archive / Unarchive — Hide without deleting
- Delete — Permanently remove a session
- Show Archived — Toggle archived session visibility
For interactive dialogue, agent switching, and conversation management, see Conversation.
Command Palette
Press Cmd/Ctrl + K anywhere in the app to open the Command Palette — a keyboard-driven shortcut to every major action.
Navigation
| Command | Destination |
|---|---|
| Cowork | Go to the Cowork workspace |
| Conversation | Go to the Conversation tab |
| Agent | Go to Agent Control |
| Jobs | Go to Jobs |
| Skills | Go to Skills |
| Programs | Go to Programs |
| Credentials | Go to Credentials |
| Settings | Open Settings |
Actions
| Command | Action |
|---|---|
| New Task | Create a new session (Cmd/Ctrl + N) |
| Theme: Light / Dark / System | Switch app theme instantly |
How to Use
- Press
Cmd/Ctrl + Kto open the palette - Start typing to filter commands
- Use arrow keys to navigate results
- Press Enter to execute the selected command
- Press Escape to close without selecting
Interactive Responses
Sometimes the agent needs your input to continue:
Action Requests
A blue "Response Required" widget appears when the agent asks a question:
- Choice buttons — If the agent provides options, click one to respond
- Free-text input — Type a custom answer if none of the options fit
- Submit — Send your response to unblock the agent
User Action Required
A yellow banner appears when the agent needs you to take an action outside the app (e.g., complete a step in a browser window or approve a permission).
Next Steps
- Conversation — Interactive dialogue, agent switching, and conversation management
- Agent Control — Manage the background agent process
- Desktop Tools — Skills, Jobs, Programs, and Credentials