AI Agent (Sub-Agent)
Eidolon's AI Agent feature goes beyond chatting — it can operate your computer, search the web, and execute complex multi-step tasks on your behalf.
The AI Agent feature is only available to Genesis tier subscribers. See Subscription & Benefits.
What Is an AI Agent?
AI Agent is Eidolon's "action" mode. When you give AI a task, it:
- Decomposes the task: Breaks complex tasks into multiple steps
- Executes step by step: Completes each operation in order
- Reports in real-time: Shows execution progress and logs in the chat
- Returns results: Reports the outcome when the task is complete
How to Use
Enable AI Agent
- Go to Settings → Preferences
- Find the "AI Agent" toggle and enable it (requires Genesis tier)
Assign Tasks to AI
Tell AI what you want it to do in the chat:
- "Search for the recent weather trends"
- "Open the browser and check tomorrow's flights"
- "Help me organize the files on my desktop"
Execution Process
When the AI Agent executes a task, the chat shows:
| Info | Description |
|---|---|
| Execution To-Dos | List of steps currently being executed |
| Phase | Current execution phase |
| Tool Calls | Tool name and parameters being invoked |
| Results | Execution result for each step |
| Thought Process | AI's reasoning and decision-making |
Capabilities
The AI Agent can perform operations including:
Information Search
- Web search (via Tavily search engine)
- Knowledge base queries (e.g. game guides)
Computer Operations
- Browser automation (navigation, clicking, typing)
- Screen capture and analysis
- File operations
Tool Integration
- MCP (Model Context Protocol) tool calls
- Custom tool extensions
Execution Logs
Each AI Agent task generates detailed execution logs:
| Log Type | Description |
|---|---|
| started | Task began |
| thought | AI's thinking process |
| tool_call | Invoking a tool |
| tool_result | Tool returned a result |
| completed | Task completed |
| error | Execution error |
| give_up | AI gave up the task |
| user_input | Needs user confirmation |
Safety & Permissions
The AI Agent follows safety principles during execution:
- Sensitive operations require confirmation: Destructive actions like deletion or modification ask for your consent first
- Operations are traceable: All execution steps are logged
- Can be interrupted anytime: You can tell AI to stop at any point
The AI Agent has the ability to operate your computer. Use it in a trusted environment only. Do not ask AI to perform operations that could cause data loss.
Common Issues
AI Agent execution failed?
- Check your network connection
- Make sure the task description is clear
- Try simplifying the task and executing it step by step
AI Agent too slow?
- Complex tasks require more steps — be patient
- You can tell AI to skip certain steps
How to disable AI Agent?
Go to Settings → Preferences and turn off the "AI Agent" toggle.