Chat Commands

When chatting with an agent via supyagent chat, you can use slash commands to manage sessions, inspect tools, control debugging, and more. Type /help to see the full list at any time.

Command Reference

Command Details

/help

/help

Displays all available slash commands with brief descriptions. This is the same list shown above.

/image

/image /path/to/screenshot.png What's wrong with this UI?
/image ./diagram.png

Sends an image to the agent along with an optional text message. The image is encoded and sent as multimodal content. Supports PNG, JPG, GIF, and WebP formats.

The path argument supports tab completion in the terminal.

/tools

/tools

Lists all tools available to the agent, grouped by source:

• supypower -- Tools from Python scripts in powers/
• remote -- Service tools from connected integrations (Gmail, Slack, etc.)
• native -- Built-in tools (delegation, process management, credentials, memory)

Each tool shows its name and a brief description.

/creds

/creds list              # List stored credentials for this agent
/creds set API_KEY       # Set a credential value (prompted securely)
/creds delete API_KEY    # Remove a stored credential

Manages per-agent credentials stored encrypted in .supyagent/credentials/{agent}.enc. Credentials are made available as environment variables to tool execution.

/sessions

/sessions

Lists all sessions for the current agent with:

• Session ID
• Creation date
• Message count
• Display title (if set with /rename)

/session

/session a1b2c3d4

Switches to a different session. The conversation history, context summary, and all state are loaded from the target session. The previous session is preserved.

/new

/new

Starts a fresh session. The current session is preserved and can be resumed later with /session.

/delete

/delete              # Delete the current session
/delete a1b2c3d4     # Delete a specific session

Permanently removes a session and its associated files (JSONL, summary, media).

/rename

/rename My refactoring project

Sets a human-readable title for the current session. Titles appear in the /sessions list for easier identification.

/history

/history         # Show last 10 messages
/history 5       # Show last 5 messages
/history 20      # Show last 20 messages

Displays recent conversation messages with their role (user, assistant, tool) and content. Tool calls show the tool name and a summary of the result.

/context

/context

Shows the current state of context management:

• Messages since last summary -- Count and threshold (e.g., 12 / 30)
• Total tokens -- Count and threshold (e.g., 45,000 / 128,000)
• Progress bars -- Visual percentage toward each trigger
• Will trigger -- Whether summarization will trigger on the next turn

This is useful for understanding how close you are to a summarization event and how much of the context window is being used.

/tokens

/tokens

Toggles the display of token usage after each response. When enabled, you see the input tokens, output tokens, and total for each LLM call.

/debug

/debug           # Toggle debug mode
/debug on        # Enable debug mode
/debug off       # Disable debug mode

Debug mode shows verbose output including:

• Raw LLM requests and responses
• Tool call arguments and results
• Context window calculations
• Retry and fallback events

/summarize

/summarize

Forces context summarization immediately, regardless of whether the thresholds have been reached. Useful when you know the conversation is about to shift topics and you want to capture the current state.

The generated summary is a structured state snapshot covering files modified, key decisions, important values, current state, and pending tasks.

/export

/export                     # Export to default filename
/export conversation.md     # Export to specific file

Exports the current conversation to a Markdown file. The export includes:

• User messages
• Assistant responses (rendered as Markdown)
• Tool calls with names and results

The path argument supports tab completion.

/model

/model                          # Show current model
/model openai/gpt-4o           # Switch to GPT-4o
/model anthropic/claude-sonnet-4-5-20250929  # Switch to Claude

Without arguments, shows the current model. With an argument, switches to the specified model for the remainder of the session. This does not modify the YAML config file.

/reload

/reload

Reloads tools from supypowers and any connected service integrations. Useful after:

• Creating new tool scripts in powers/
• Connecting new integrations via supyagent connect
• Modifying existing tool scripts

Displays the new tool count after reloading.

/clear

/clear

Clears the terminal screen. Does not affect conversation history.

/quit

/quit

Exits the chat session. The session is automatically saved and can be resumed later.

Tab Completion

The chat input supports tab completion for:

• Slash commands -- Type / and press Tab to see all commands with descriptions
• File paths -- After /image or /export, Tab completes file paths
• Credential actions -- After /creds, Tab completes list, set, delete
• Debug options -- After /debug, Tab completes on, off
• @-path references -- Type @ followed by a path fragment and Tab to complete

Input Features

• Command history -- Up/Down arrows navigate through previous inputs
• History search -- Ctrl+R searches through input history
• Syntax highlighting -- Slash commands are highlighted in green, @-paths in blue
• Auto-suggest -- Ghost text from history appears as you type

Related

• Sessions -- Session persistence and context management details
• Quickstart -- Getting started with chat
• Tool Permissions -- What tools are available