Chat Thread

The main area displays your conversation as a scrolling message thread:

• Your messages — Shown with a "User" badge
• Claude's responses — Shown with a "Claude" badge, rendered as full Markdown with syntax highlighting
• Tool call cards — Inline cards showing Claude's actions (file reads, writes, bash commands, etc.)
• Plan cards — Interactive cards for reviewing and approving plans
• Question cards — Multi-select forms when Claude needs your input
• Subagent cards — Nested agent activities with expandable timelines

Markdown Rendering

Claude's responses support:

• Full GitHub-flavored Markdown
• Syntax-highlighted code blocks with a copy button
• Mermaid diagrams (rendered as interactive SVG)
• File path linking — paths like src/main/index.ts:42 become clickable links that open in the File Viewer

File Path Linking

When Claude mentions file paths in responses, Pandev automatically detects them and makes them clickable. Clicking a file path:

• Opens the file in the File Viewer panel
• If a line number is included (e.g., file.ts:42), scrolls to that line

Supported path formats:

• Absolute: /Users/you/project/src/file.ts
• Relative: src/main/index.ts, ./lib/utils.ts
• With line numbers: src/file.ts:42

Chat Input

The input area at the bottom includes:

Text Input

• Type your message and press Enter to send
• Supports multi-line input
• Drag and drop images to attach them
• Paste images from clipboard (Cmd+V)

Model Selector

A dropdown next to the input lets you switch Claude models mid-conversation. Available models depend on your API access.

Interaction Mode Selector

Choose the permission level for the session:

• Default — Claude asks for permission before actions
• Accept Edits — Claude can read and write files without asking
• Plan — Claude must propose plans before making changes
• Bypass Permissions — Claude can do everything without asking

Abort Button

While Claude is thinking, the send button becomes an Abort button. Click it or press Escape to stop the current operation.

Tool Call Cards

When Claude uses tools (reading files, running commands, editing code), each action appears as a collapsible card:

• Header — Tool icon, tool name, and status indicator
• Timer — Shows elapsed time while running
• Status — Spinning dots (running), green check (success), red X (error)
• Input — Click to expand and see what Claude passed to the tool
• Output — The tool's response or error message

Common tool types you'll see:

Message Queue

If you send messages while Claude is busy (status is "thinking"), they're added to a message queue:

• Queued messages appear above the input with a preview
• Each has a remove button (X) to cancel it
• Messages send automatically in order when Claude becomes idle
• The queue flushes with a 500ms delay between messages

Turn Stats

Optionally shown below each Claude response (enable in Settings > General > Turn Stats):

• Cost — Estimated API cost in USD
• Duration — How long the turn took
• Turns — Number of back-and-forth exchanges
• Tokens — Token usage breakdown by model

Drag and Drop

You can drag files onto the chat area:

• Image files — Attached to your next message (Claude can see and analyze images)
• Visual feedback shows a drop zone overlay when dragging