AI Agent

CommandLane's AI Agent is a conversational assistant with full access to your knowledge base. Ask questions, create entries, search your notes, run commands, and more — all through natural language.

Opening the Agent

Method	How
Mode switch	Press Shift+Tab in the capture window
Quick question	Type q (q + space) from any mode
Mode selector	Hover the mode icon (left of input) and select Agent

What You Can Do

Search and Ask Questions

The agent can search your knowledge base semantically and by keyword. Ask it anything about your captured notes, tasks, and decisions.

What did I decide about the database migration?

Summarize my open tasks for the Acme project

Create and Update Entries

Create tasks, notes, and other entries through conversation. The agent classifies them automatically.

Create a task: review the Q2 budget proposal by Friday

Update my meeting notes from yesterday — add the action items we discussed

Set Reminders

Ask the agent to remind you about something at a specific time.

Remind me to follow up with Sarah tomorrow at 10am

Run Shell Commands

The agent can execute shell commands on your machine. Sensitive commands require your approval before running.

Show me the git log for the last week

Search the Web

When connected to a supported AI provider (OpenAI or Anthropic), the agent can search the web to answer questions.

What are the latest Next.js 16 features?

File Operations

Read, write, and manage files in your workspace directory.

Read the contents of config.yaml

Create a new file called notes.md with today's meeting summary

Model Selection

Use the model dropdown in the footer bar to choose your AI model. Available models depend on which providers you've configured in Dashboard > Settings > AI Settings.

Supported providers:

• OpenAI — GPT-4o, GPT-4o-mini, and more
• Anthropic — Claude Sonnet, Claude Haiku
• Ollama — Any locally running model

@Mention Subagents

Type @ followed by an agent name to invoke a specialized agent. Ghost-text autocomplete helps you discover available agents.

@researcher Find recent papers on transformer architectures

Slash Commands

Type / to see available commands within the agent:

Command	Description
/resume	List and resume past conversations
/resume 20	Show the 20 most recent conversations
/snippet	Insert a saved text snippet

Sessions

Every agent conversation is automatically saved as a session. You can:

• Resume later: Type /resume to pick up where you left off
• Browse sessions: Open Dashboard > Sessions to see all past conversations
• Full context: Message history and conversation context are preserved

Approval Flow

When the agent wants to perform a sensitive action (deleting entries, running certain shell commands, modifying files), it asks for your approval first. You can:

• Approve the action
• Dismiss to skip it
• Grant session-level permission so the agent doesn't ask again for the same type of action during the current session

Tips

Scenario	Recommendation
Quick thought or task	Use Capture mode
One-off question	Use q prefix from Capture mode
Multi-turn planning or brainstorming	Switch to Agent mode
Finding a specific entry	Type in Capture mode for instant search
Complex knowledge base queries	Use Agent for semantic search