Interactive Chat

Launch

yeepilot run

(or simply yeepilot)

Interaction Model

You describe a goal. YeePilot analyzes context, proposes actions/commands, and asks for approval based on your active security/autonomy settings.

Essential Slash Commands

Planning and autonomy

• /plan on|off|status
• /plan <task>
• /autonomy status|high|guarded|strict_review|ultra

Reasoning and tokens

• /think on|off|low|medium|high
• /tokenmode normal|saver

Safety and execution

• /mode strict|moderate|permissive
• /sandbox on|off|status|net|nonet

Session management

• /new
• /rename <name>
• /sessions
• /history
• /history search <query>
• /status
• /context
• /clear
• /copy
• /export

File context and live output

• /file <path>
• /files
• /unfile <path>
• /upload
• /livestream [on|off|status|scope <all|files|off>]
• /sidecar

Vault and secure access

• /vault ...
• /unlock [method]
• /lock
• /connect <entry>
• /portable export|import <bundle.ypbundle>

UI/help

• /model
• /setup
• /language
• /theme
• /keybinds
• /help
• /exit

Note: /agent still works as a legacy alias, but default runtime is already agent mode.

Approval Flow

Typical cycle:

1. prompt
2. plan/proposal
3. risk visibility
4. approval
5. execution
6. verification
7. summary/review

For multi-step work, repeat until completion or bounded stop conditions.

Practical Usage Pattern

1. Start with /plan <task> for non-trivial work.
2. Keep /autonomy guarded or /autonomy strict_review for production-sensitive environments.
3. Use /tokenmode saver for long sessions.
4. Keep relevant files attached with /file.
5. Use /export for audit/report artifacts.