Type / in an interactive session to autocomplete commands. Run /help to see the commands loaded in the current session, or /help <command> for detailed usage. Skills and MCP servers can add commands at runtime, so they are not listed individually here.
#Chat and models
| Command |
Purpose |
/help [command] |
List available commands or show detailed usage for one command, including /ps. |
/clear, /new |
Start a clean chat without deleting saved sessions. |
/model [name] |
Open the model picker or switch to a specific model; Flux switches provider when needed. |
/effort [low|medium|high|max] |
Show or set reasoning effort. |
/fast, /deep |
Switch to the current provider's fast or most capable model shortcut. |
/compact |
Summarize older conversation content to free context-window space. |
/context |
Show a token-use breakdown for the system prompt, tools, MCP tools, skills, and messages. |
/export |
Write the current conversation, including tool activity, to a timestamped Markdown file in the working directory. |
/ps <question> |
Run a brief independent side query with read-only conversation context, without interrupting the active main task. It cannot use tools or plugins. |
#Sessions and workflow
| Command |
Purpose |
/history [index] |
List saved sessions or load one by its displayed index. |
/search <query> |
Search saved session messages by keyword. |
/name <title> |
Give the current session a descriptive name. |
/branch [name] |
List branches in the current session lineage, or fork the current conversation into a named branch. |
/plan [clear] |
Show the active plan or delete it. Use no plan in a prompt to skip automatic planning. |
/steer <directive> |
Add durable direction for later turns. Use /steer list, /steer remove <number>, or /steer clear to manage directives. |
/loop |
Show the active loop. /loop <interval> <prompt> repeats an agent prompt; intervals support s, m, and h. /loop <prompt> defaults to ten minutes, and /loop stop cancels it. |
/schedule |
List saved tasks. Use /schedule add "<five-field cron>" <prompt>, then enable, disable, or remove <id> to manage them. |
/watch <path|glob> <command> |
Run a shell command after matching files change. /watch stop cancels it; glob patterns are supported. |
/undo, /undo list |
Revert the most recent tracked file change or inspect the current in-memory undo stack. |
#Memory and skills
| Command |
Purpose |
/memory, /memory list |
List project-scoped persistent memories. |
/memory add <text> |
Save a persistent memory. |
/memory delete <name>, /memory rm <name> |
Delete one memory by its listed name. |
/memory clear, /memory path |
Delete all project memories or show their storage directory. |
/remember <text> |
Shortcut for /memory add <text>. |
/forget [name] |
List memories when called without an argument, or delete a memory by name. |
/skills, /skills list |
List bundled, global, and project skills. |
/skills info <name> |
Show a skill's origin, invocation mode, tool policy, path, and usage statistics. |
/skills add <path|url> |
Validate and install a skill from a local SKILL.md or trusted URL. Restart Flux to activate it. |
/skills marketplace [number] |
List available marketplace skills or install a numbered entry. |
/skills remove <name>, /skills rm <name> |
Remove an installed global skill. Restart Flux to apply the change. |
/<skill-name> [arguments] |
Invoke an installed user-callable skill. The exact command name comes from the skill's slug. |
#Diagnostics and exit
| Command |
Purpose |
/debug |
Show conversation, plan, sandbox, token, plugin, MCP, undo, and memory state. |
/changelog |
Show release notes from the latest update. |
/exit |
Save the current session and exit Flux. |
MCP prompts appear as additional slash commands supplied by their connected server. Project skills and user-installed skills likewise add commands after Flux restarts; use /help or type / to see what is available in the current session.