Slash Commands

Slash commands are typed directly in the input during a session. They control context, models, permissions, and session lifecycle.

Command Reference

Command	Aliases	Description
/help		Show commands and shortcuts
/model		Switch between configured models
/key		Add, replace, remove, or import a provider API key
/tier		Configure sub-agent model tiers (high/medium/low)
/permission		Set permission mode (read_only / reversible / yolo)
/summarize		Interactively summarize older context
/summarize_hint		Configure the two-tier summarize hints (on/off, trigger levels)
/compact		Full context reset with continuation summary
/rewind	/undo	Rewind to a previous turn (reverts conversation + files)
/review		Review code changes (base branch / uncommitted / commit / custom)
/session	/resume	Resume a previous session
/new		Start a new session
/fork		Fork current session into a new branch
/rename		Rename current session
/shells		View and stop background shells
/usage	/context	Show this session’s token usage
/stat		Show all-time token statistics
/autoupdate		Toggle background update checks
/autocopy		Toggle copy-on-select (auto-copy a text selection)
/skills		Enable/disable skills (checkbox picker)
/mcp		Manage MCP servers and list tools
/agents		Toggle the agents panel
/todos		Toggle the todo panel
/theme		Set theme mode (auto / light / dark)
/diff		Set write/edit diff display (compact / full)
/codex		OpenAI ChatGPT OAuth login
/copilot		GitHub Copilot login
/hooks		Manage registered hooks
/copy		Copy the agent’s most recent text response
/raw	/md	Toggle markdown raw/rendered mode
/quit	/exit	Exit the application

Each enabled, user-invocable skill also appears as its own /<skill-name> command. Type / to see the full list — built-in commands and skills together.

Context Commands

/summarize

Opens an interactive range picker:

1. Select the start turn
2. Select the end turn
3. Optionally provide a focus prompt — instructions about what to preserve

The selected range is converted to context IDs and summarized. Key decisions and findings are preserved; the rest is discarded.

/summarize

See Context Management for details.

/compact

Full context reset with a continuation summary. Optionally provide instructions:

/compact
/compact Preserve the DB schema decisions

After compact, the agent starts with a fresh context window containing only the continuation summary and AGENTS.md files.

/summarize_hint

Configure the two automatic hints that nudge the agent to summarize as context fills up. Toggle them on/off, or set custom trigger percentages:

/summarize_hint on
/summarize_hint off
/summarize_hint 50 75

The two integers are the Level 1 and Level 2 trigger percentages and must satisfy 0 < level1 < level2 < 85. The setting persists. See Context Management.

/rewind

Roll back to a previous turn. Opens a picker showing turn history. Reverts both the conversation and file system changes made after that turn.

/rewind

File revert uses tracked mutations (edits, writes, bash mkdir/cp/mv) with conflict detection — if a file was modified externally after the agent changed it, the rewind skips that file and reports the conflict.

/review

Run a code review. Opens a picker to choose what to review:

/review

• Against a base branch — review the diff vs. a base branch
• Uncommitted changes — review the current working-tree changes
• A commit — review a specific commit by SHA
• Custom instructions — provide your own review focus

You can also pass instructions inline: /review check for SQL injection risks. The picker supports adding extra instructions (Tab) to any option.

Model Commands

/model

Opens a hierarchical picker showing all configured providers and models. Select one to switch immediately.

For managed providers with missing API keys (Kimi, MiniMax, GLM, DeepSeek, Xiaomi, Qwen), selecting a model can prompt you to paste or import the key on the spot.

/key

Manage a provider’s API key directly — add, replace, remove, or import a detected external env var. Opens a picker of provider endpoints, then an action menu. Keys are stored in ~/.fermi/.env (0600). This is the quickest way to rotate or fix a key without re-running fermi init.

/tier

Configure which models sub-agents use at each tier level (high, medium, low). When the agent spawns a sub-agent with model_level="low", it uses the model you assigned to the low tier.

Permission & Safety

/permission

Set the permission mode for tool execution:

Mode	Behavior
read_only	Only read tools auto-allowed; all writes require approval
reversible	Read + reversible writes auto-allowed
yolo	Everything auto-allowed except catastrophic operations

/autoupdate

Toggle background update checks:

/autoupdate on
/autoupdate off

When enabled, Fermi checks GitHub Releases in the background, stages patch/minor updates automatically, and applies them on the next restart.

/hooks

Show all registered hooks and their configuration. Hooks are shell commands that run in response to events (PreToolUse, PostToolUse, UserPromptSubmit, etc.).

Session Commands

/session

Resume a previous session. Shows a list of saved sessions with timestamps. Select one to restore its full conversation state.

/session
/session <id>

/new

Start a new session. The current session is auto-saved first.

/fork

Fork the current session into a new branch — creates a copy of the current state that you can take in a different direction.

/rename

Give the current session a descriptive name for easier identification in the session list.

Other Commands

/skills

Opens a checkbox picker to enable or disable installed skills. Skills are loaded from multiple roots — global (~/.fermi/skills/), project (<project>/.fermi/skills/), and the per-project store — and a toggle here reloads them immediately.

/mcp

Connects configured MCP servers and lists discovered tools. Useful as a health check before starting work.

/agents

Toggles the agents panel — the main agent and any running sub-agents with their status.

/todos

Toggles the todo panel, which shows the agent’s current plan/task list.

/theme

Set the TUI theme mode: auto, light, or dark. Persisted to settings.json (theme_mode).

/diff

Set how write/edit diffs are rendered: compact or full. Persisted to settings.json (diff_display).

/codex / /copilot

OAuth login flows for OpenAI (ChatGPT) and GitHub Copilot respectively.

/copy

Copy the agent’s most recent text response to the system clipboard.

/raw

Toggle between rendered markdown and raw markdown display. Also available as /md.

/shells

View tracked background shells and stop them. Background shells come from bash_background or from a bash call that ran past its timeout and was handed off to the background.

/usage and /stat

/usage (alias /context) shows the current session’s token usage — how much of the context budget is consumed and the breakdown by category. /stat shows your all-time token statistics across every session.

/autocopy

Toggle copy-on-select. When on, selecting text with the mouse automatically copies it to the clipboard (with a brief toast), so you don’t need a separate copy step.

/autocopy on
/autocopy off

Keyboard Shortcuts

Shortcut	Action
Enter	Send message
Option+Enter / Ctrl+N	Insert newline
↑ / ↓	Browse prompt history
Ctrl+Q	Cycle permission mode
Ctrl+V	Paste image
Ctrl+G	Toggle markdown raw view
PageUp / PageDown	Scroll half page
Opt+← / →	Switch between agent tabs
Ctrl+X	Kill all sub-agents
Ctrl+K	Kill all background shells
Alt+Backspace / Ctrl+W	Delete previous word
Cmd+Delete	Delete to line start (Ghostty/kitty protocol)
Ctrl+C	Cancel / Exit
@filename	Attach file