other commands
Documentation for kas monitor, kas instance, kas audit, kas tmux, kas status, kas reset, kas debug, kas version, and kas check.
kas monitor
Monitor the kasmos daemon event stream in real time via Server-Sent Events (SSE).
kas monitor [--socket <path>] [--repo <path>] [--plan <slug>] [--json]
Alias: kas mon
# stream all daemon events
kas monitor
# filter to a specific repo
kas monitor --repo /home/user/myproject
# filter to a specific plan slug
kas monitor --plan my-feature
# pipe raw JSON to jq
kas monitor --json | jq .
| flag | default | description |
|---|---|---|
--socket | auto-detected | path to the daemon unix domain socket |
--repo | — | filter events to a specific repo path |
--plan | — | filter events to a specific plan slug |
--json | false | output raw JSON (disables ANSI color formatting) |
By default, events are formatted with ANSI color codes. ctrl-c stops the stream.
monitor status
Show a one-shot snapshot of daemon state: registered repos and active agents.
kas monitor status [--socket <path>]
kas monitor status
kas instance
Manage agent instances tracked by kasmos.
kas instance <subcommand> [flags]
Alias: kas i
instance list
List all tracked agent instances.
kas instance list [--format text|json] [--status <status>]
# text table (default)
kas instance list
# JSON output for scripting
kas instance list --format json
# only running instances
kas instance list --status running
| flag | default | description |
|---|---|---|
--format | text | output format: text or json |
--status | — | filter by status: running, ready, loading, paused |
instance kill
Pause an agent instance (safe kill): kills the tmux session and removes the worktree but preserves the branch. The instance is marked paused in state.
kas instance kill <title>
kas instance kill "my-agent"
Requires a clean worktree (no uncommitted changes). Use an exact or unique substring match for the title.
instance pause
Pause an agent instance: same as kill but more explicit. Kills the tmux session and removes the worktree while preserving the branch.
kas instance pause <title>
instance resume
Resume a paused agent instance by recreating the worktree and starting a new tmux session.
kas instance resume <title>
kas instance resume "my-agent"
Re-adds the git worktree from the stored branch name, reconstructs the original program command (including env vars like KASMOS_MANAGED=1, KASMOS_TASK, KASMOS_WAVE, KASMOS_PEERS), and starts a new kas_-prefixed tmux session.
instance send
Send a prompt to a running agent instance via tmux send-keys.
kas instance send <title> <prompt>
kas instance send "my-agent" "please implement the feature"
The instance must not be in paused state.
instance status
Show a summary of instance counts by status bucket.
kas instance status
kas instance status
# STATE COUNT
# running 2
# ready 1
# paused 0
# killed 0
kas audit
Query audit events recorded by the kasmos audit logger.
kas audit <subcommand> [flags]
Alias: kas au
audit list
List recent audit events from the local SQLite audit log.
kas audit list [--limit <n>] [--event <kind>]
# show last 50 events (default)
kas audit list
# show last 100 events
kas audit list --limit 100
# filter to a specific event kind
kas audit list --event task_status_changed
| flag | default | description |
|---|---|---|
--limit | 50 | maximum number of rows to return |
--event | — | filter by event kind |
kas tmux
Manage orphan kas_-prefixed tmux sessions that are not tracked by kasmos.
kas tmux <subcommand>
Alias: kas tx
tmux list
List orphan tmux sessions (sessions with a kas_ prefix that are not in the kasmos instance state).
kas tmux list
kas tmux list
# NAME TITLE WINDOWS ATTACHED AGE
# kas_old-feature old-feature 1 false 2h ago
tmux adopt
Adopt an orphan tmux session as a managed kasmos instance.
kas tmux adopt <session> <title>
kas tmux adopt kas_old-feature "old-feature"
The session must be a kas_-prefixed orphan (not already tracked). Creates a new instance record with status ready.
tmux kill
Kill an orphan tmux session.
kas tmux kill <session>
kas tmux kill kas_old-feature
The session must be an orphan (not tracked by kasmos). Use kas tmux list to see eligible sessions.
kas status
Show a unified overview of active tasks, agent instances, and orphan tmux sessions.
kas status [--json]
Alias: kas st
# human-readable text (default)
kas status
# machine-readable JSON
kas status --json
| flag | description |
|---|---|
--json | output as JSON ({tasks, instances, orphan_sessions}) |
The text output includes a hints: section with suggested follow-up commands when relevant (e.g. kas task implement for ready tasks, kas instance resume for paused instances).
kas reset
Reset all stored instances, clean up tmux sessions and git worktrees, and stop the daemon.
kas reset
kas reset
# Storage has been reset successfully
# Tmux sessions have been cleaned up
# Worktrees have been cleaned up
# daemon has been stopped
This command is destructive — all instance state is deleted and all kas_-prefixed tmux sessions are killed. Use it to start fresh.
kas debug
Print config paths and the current configuration as JSON.
kas debug
kas debug
# Config: /home/user/.config/kasmos/config.toml
# {
# "defaultProgram": "claude",
# ...
# }
kas version
Print the kasmos version number.
kas version
kas version
# kas version v2.0.0-alpha
# https://github.com/kastheco/kasmos/releases/tag/v2.0.0-alpha
kas check
Audit skill sync health across all configured harnesses.
kas check [-v]
# summary view
kas check
# verbose: show per-skill status for each harness
kas check -v
| flag | description |
|---|---|
-v, --verbose | show per-skill detail for each harness |
Checks two skill layers:
- Global skills —
~/.agents/skills/→ harness global directories (e.g.~/.claude/skills/) - Project skills —
.agents/skills/→ harness project directories
Each skill is reported as one of: synced (✓), skipped (⊘), missing (✗), orphan (✗), broken (✗), or copy (≈).
The command exits with code 0 when all checks pass (100% health) and code 1 otherwise. Remediation hints are printed when issues are found.
# typical healthy output
# Global skills (~/.agents/skills):
# claude 3 synced 0 skipped 0 missing 0 orphan
# opencode 3 synced 0 skipped 0 missing 0 orphan
#
# Health: 6/6 OK (100%)