Your First Session

Start your first session and see what Lucid AI can do

Now that you're all set up, kick off your first Lucid AI session! This guide will walk you through the new session interface and help you understand the best ways to interact with Lucid AI.

Understanding the Lucid AI Session Page

When you start a new session, you'll see two primary modes: Ask and Agent.

Ask Mode

Ask mode is for code research, browsing documentation, and planning. Use it to:

• Ask questions about how your code works
• Explore architecture and dependencies
• Plan and scope tasks before implementation
• Generate context-rich prompts for Agent sessions

Triggering Ask Mode

From the main page, type a query in the Ask input and press Enter. See the Ask Lucid AI guide for more details.

For Ask mode from a DeepWiki page, type a query in the chat input at the bottom of the page and click Ask. This will automatically scope Lucid AI's knowledge to that repository specifically.

Agent Mode

Agent mode is Lucid AI's full autonomous mode where it can write code, run commands, browse the web, and complete complex tasks end-to-end. Use Agent mode when you're ready to:

• Implement features or fix bugs
• Create pull requests
• Run tests and debug issues
• Perform multi-step tasks that require code changes

Triggering Agent Mode

You can trigger Agent mode from the main page or from an Ask Lucid AI session.

For tasks that are not fully scoped, we recommend:

• Start with Ask mode to plan out the task
• Construct a Lucid AI Prompt, which will draw from your Ask session to create a scoped plan
• Click Send to Lucid AI to move to Agent mode and execute the task

For Agent mode from the main page, toggle to Agent mode and select the repository/repositories you want to work with.

When starting an Agent session, you'll configure a few options: selecting a Repository and selecting an Agent.

Selecting a Repository

Select the repository you want Lucid AI to work with. Click the repository selector to see all repositories that have been added to Lucid AI's machine.

When you select a repository, Lucid AI:

• Has access to your codebase and can make changes
• Uses the correct branch as a starting point
• Can create pull requests to the right repository

Selecting an Agent

You can choose which agent configuration Lucid AI uses for your session. Different agents may have different capabilities or be optimized for specific types of tasks.

Currently, we have a default agent that works well for most tasks, and a data analyst agent named Dana that is optimized for data analysis tasks.

Using @ Mentions

When you type @ in the chat input, you'll see a dropdown of available mentions:

• @Repos - Reference a specific repository
• @Files - Reference a specific file in your codebase
• @Macros - Reference a macro for a Knowledge entry
• @Playbooks - Reference a team or community playbook, which are detailed prompt templates that can be used to guide Lucid AI's behavior.
• @Secrets - Reference a specific secret (e.g. API keys, credentials, etc.) from Lucid AI's session manager

Scoping Your First Session

Start with tasks that have clear success criteria and provide Lucid AI with the context it needs — just as you would when handing off work to a teammate. As you get comfortable, try progressively more complex tasks. We've seen users work with Lucid AI on everything from fixing small bugs to targeted refactors to large-scale migrations and building entire features from scratch.

First-time Prompt Ideas

After Your Session

Once Lucid AI finishes, check out Session Insights — you'll get a timeline of what happened, actionable feedback, and an improved prompt you can use for similar tasks in the future.

Next Steps

Once you're comfortable with basic sessions, explore these resources to get more out of Lucid AI: