Chatting

Start a chat from the menu (Start a new chat) or with:

kunji chat

Type your message and press Enter. The answer streams in as it’s written, then a grey line shows the model, tokens, cost and time:

› what is the difference between a process and a thread?
  A process is a running program with its own memory. A thread is a path of
  execution inside a process; threads of the same process share memory…
  claude-sonnet · 21 in · 148 out · $0.0023 · 2.1s

Writing messages

Key	Action
Enter	Send
Alt+Enter or Ctrl+J	New line (for longer messages or pasted code)
↑ / ↓	Bring back a message you typed earlier
Tab	Complete commands, model names, prompt names and file paths
Ctrl+C	Stop the answer that’s being written (what’s written so far is kept)
Ctrl+D	Leave the chat

Slash commands

Type / to see them all. The most useful ones:

Command	What it does
/model provider:model	Switch model, keep the conversation (more)
/file app.py	Attach a file to your next message (more)
@explain …	Use a saved prompt (more)
/cost	What this chat and today have cost
/retry	Get a new answer to your last message
/new	Start a fresh conversation
/save notes.md	Export the chat as Markdown

Full list: Chat commands.

System prompts

A system prompt sets how the model should behave for the whole chat.

kunji chat -s "You are a senior Python developer. Be concise."

Inside a chat: /system Answer in Hindi. and /system off to clear it. To use one for every new chat, set it in Settings → Default system prompt or:

kunji config set chat.system_prompt "Be concise."

Starting a chat with a specific model

kunji chat -m openai:gpt-4o-mini

Models are written as provider:model. See Switch and pick models.

Continuing later

Chats are saved automatically. Continue the last one with kunji chat -c, or pick any saved chat; see Chat history.

One-off questions

For a single question without opening a chat, use kunji ask:

kunji ask "How do I undo the last git commit?"

See Pipes and scripts.