Skip to content

Commands

clam add

Create a new profile.

clam add <name> [--inherit]
Argument Description
<name> Profile name (letters, numbers, hyphens, underscores only)
--inherit Symlink global Claude Code config (~/.claude/) into the profile

Creates a profile directory at ~/.clam/profiles/<name>/ with 0700 permissions. If this is the first profile, it is auto-activated.

With --inherit, shared config files from ~/.claude/ are symlinked into the profile. Session-specific items (auth, history, cache, telemetry) are skipped to maintain isolation.

# Basic profile
clam add personal

# Profile with shared config
clam add work --inherit

clam list

List all profiles.

clam list [--json] [--names]
Argument Description
--json Output as JSON array
--names Print only profile names, one per line (for shell scripts)

Text output marks the active profile with *:

* work
  personal

JSON output:

[
  { "name": "work", "active": true },
  { "name": "personal", "active": false }
]

clam use

Switch the active profile.

clam use <name>
Argument Description
<name> Profile name to activate

Updates ~/.clam/state.toml to point at the given profile. Does not affect running Claude Code sessions — only new sessions pick up the change.

clam current

Print the active profile name.

clam current

Prints the name to stdout and exits 0. If no profile is active, prints nothing and exits 1.

Useful for scripting:

if profile=$(clam current); then
  echo "Active: $profile"
else
  echo "No active profile"
fi

clam remove

Remove a profile.

clam remove <name> [--force]
Argument Description
<name> Profile name to remove
--force Skip confirmation prompt

Deletes the profile directory and removes it from the registry. Cannot remove the currently active profile — switch to another profile first.

Without --force, prompts for confirmation. In non-interactive contexts (pipes, scripts), --force is required.

# Interactive removal
clam remove old-profile

# Scripted removal
clam remove old-profile --force

clam status

Show details about the active profile.

clam status [--json]
Argument Description
--json Output as JSON object

Text output:

Profile: work
Path:    /Users/you/.clam/profiles/work

If no profile is active, prints an error to stderr and exits 1.

clam run

Launch a command with the profile's environment.

clam run [--profile <name>] [-- <command> [args...]]
Argument Description
--profile <name> Use a specific profile without switching the active one
-- <command> Command to run (defaults to claude)

Sets CLAUDE_CONFIG_DIR to the profile directory, sets CLAM_PROFILE to the profile name, and strips all CLAUDE_* and ANTHROPIC_* env vars from the parent shell. Then exec()s into the target command (replacing the clam process).

# Launch Claude Code with active profile
clam run

# Launch with a specific profile
clam run --profile work

# Pass arguments to claude
clam run -- --help

# Run any command with profile environment
clam run -- env | grep CLAUDE

clam env

Print shell environment exports.

clam env [--json]
Argument Description
--json Output as JSON object instead of shell exports

Outputs export and unset statements suitable for eval:

eval "$(clam env)"

This unsets any existing CLAUDE_* and ANTHROPIC_* variables, then exports CLAUDE_CONFIG_DIR and CLAM_PROFILE for the active profile.

clam completions

Generate shell completion scripts.

clam completions <shell>
Argument Description
<shell> One of: bash, zsh, fish 
# Bash
clam completions bash > ~/.local/share/bash-completion/completions/clam

# Zsh
clam completions zsh > ~/.zsh/completions/_clam

# Fish
clam completions fish > ~/.config/fish/completions/clam.fish

clam repair

Repair legacy profiles that share session data with the global ~/.claude/.

clam repair [--dry-run] [--profile <name>] [--force]
Argument Description
--dry-run Show the plan without making any changes
--profile <name> Repair only a specific profile (default: all)
--force Skip the confirmation prompt

Older clam versions symlinked projects/ and todos/ inside each profile to the shared ~/.claude/projects/ and ~/.claude/todos/. Because Claude Code reads session transcripts from projects/ (for /resume and the session picker) and per-session todo state from todos/, this caused sessions created in one profile to appear in every other profile.

clam repair detects these bad symlinks, converts them into real per-profile directories, and migrates owned session data in.

How ownership is decided

Each profile's .claude.json contains a projects map keyed by the absolute working directories Claude Code has opened under that profile. clam repair uses this as the source of truth:

  • If a working directory is claimed by exactly one profile, its session transcripts and matching todo files are moved into that profile.
  • If a working directory is claimed by multiple profiles (a conflict), it's left in ~/.claude/ and listed in the plan — resolve it manually.
  • Working directories claimed by no profile stay in ~/.claude/ as orphans (safe to leave, used only by direct Claude Code invocations outside clam).
# 1. Preview the plan
clam repair --dry-run

# 2. Back up first (safety net)
tar czf ~/claude-backup-$(date +%Y%m%d).tgz -C ~ .claude/projects .claude/todos

# 3. Execute
clam repair

After repair, each profile's ~/.clam/profiles/<name>/projects/ and todos/ are real directories containing only that profile's session data. clam run <profile> will show isolated session history in /resume.

Note

clam repair is idempotent and safe to re-run — profiles that are already isolated are skipped with a "Nothing to repair" message.