Claude Code

SuperClawd is a launcher for Claude Code, Anthropic's official CLI. The superclawd command starts Claude Code with your workspace's skills, agents, commands, and always-on rules already wired in — no manual setup.

Prerequisites

• Claude Code installed and on your PATH
• The SuperClawd CLI installed (npm i -g superclawd)
• This machine connected — run superclawd login (see Devices)

Launch

From any project directory, run:

superclawd

This opens Claude Code in your current directory with your SuperClawd workspace loaded. (superclawd works too — superclawd is the short alias.)

Non-interactive mode

Pass -p (print mode) to run a single prompt without the interactive UI. SuperClawd loads your workspace, hands the prompt to claude -p, prints the result, and exits — ideal for scripts, CI, and one-off questions. Your slash commands work here too (namespace them as /superclawd:<command>).

superclawd -p "Summarize the changes in the last commit"

Print mode never prompts (so it can't stop to ask), which means it always runs with permissions skipped for that run. Two flags refine a print-mode run:

• --workspace <id> — run against a specific workspace instead of your default. Print mode never shows the workspace picker, so use this to target one explicitly.
• --no-capture — skip writing learned memory for this run. Handy for behavior tests and throwaway prompts you don't want to teach from.

superclawd -p "Run the release checklist" --workspace my-team --no-capture

Resume a conversation

Pick up a previous Claude Code conversation with --resume (alias -r) — it's forwarded to Claude Code with your SuperClawd workspace freshly loaded. Pass a session id to resume that conversation directly, or run it bare to choose from a list.

superclawd --resume                 # pick from recent conversations
superclawd --resume <session-id>     # resume a specific one

Keep your computer awake

Long runs can be cut short if your machine sleeps while the terminal is unfocused. Toggle Keep Awake in superclawd config (off by default) and SuperClawd holds a wake lock for the duration of each session, releasing it automatically when the session ends.

In the session

• Skills activate automatically — when a task matches one of your workspace skills, Claude fetches and applies its instructions on demand. Only a skill's stable release is served; a skill that has never been released is skipped. See Versions & Releases.
• Experimental mode — toggle it in superclawd config (off by default) to serve the live draft of any skill you've flagged experimental, instead of its stable release. See Experimental Mode.
• Foundational rules and workspace memories are always on — your always-active guidelines and team memories are injected at the start of every session and re-applied as it grows. See Team Memories.
• Learned memories surface automatically — relevant learned memories now appear on their own on each new prompt and when a file is opened, and are recalled by meaning, not just exact words. Opt out with SUPERCLAWD_AUTO_RETRIEVE=0. See Learned Memory.
• Agents from your workspace show up as Claude Code subagents, served whole at catalog time. Like skills, only an agent's stable release is served — an agent that has never been released is skipped (no fallback), and a flagged agent serves its live draft in an experimental session. See Agent versions.
• Commands and workflows from your workspace show up as Claude Code slash commands.

Slash commands

SuperClawd adds these to Claude Code's / menu:

• /superclawd:status — service status, workspace info, and credit balance
• /superclawd:audit-work — audit the session against your active SuperClawd rules
• /superclawd:ultrathink — apply maximum extended thinking to the task at hand
• /superclawd:notifications — toggle notifications for the current session
• /superclawd:<workflow> — run one of your workspace's workflows

Permissions

By default superclawd launches Claude Code normally — with its usual approval prompts for tools, file edits, and commands.

To skip those prompts, run superclawd config and toggle Dangerously skip permissions on. It applies to future launches; turn it off anytime from the same menu.