Skip to content

CLI reference

Every Showtail command, grouped the way showtail --help shows them. Run showtail <command> --help for the authoritative, up-to-date flag list.

showtail <command> [options]

Global flags: -v, --version prints the version. Most commands accept --json for machine-readable output (noted below).

Get started

Command What it does
setup One-time guided setup: connect your AI tools and turn on automatic tracking. Flags: --off (turn tracking back off), --yes (no prompts), --json.
track [path] Track a folder as a Showtail project (creates .showtail/) and pull its already-captured work out of the inbox. Any folder works — it need not be a code repo. Flags: -p, --project <name>, --json.
ensure Make sure this project is initialized and a session is open (safe to re-run; this is what the hooks call). Flags: --json.
start Begin a new work session. Flags: -l, --label <label>, --json.
end Close the current work session. Flags: --json.

You rarely need track or start

Once a tool is connected, Showtail initializes the project and opens a session automatically the first time you work. track/start are for wiring things up by hand (or to declare a non-code folder — like a book — as a project).

Capture your work

Command What it does
log Record an event (usually a prompt) in your current session. Flags: -t, --type <type> (required), -x, --text <text> (or pipe via stdin), -f, --files <files>, --tool <tool>, -s, --session <id>, --turn <id>.
artifact <file> Snapshot a file's current state (hash, time, git commit). Flags: -s, --session <id>, --tool <tool>.

Review your trail

Command What it does
status Your current session and connected tools at a glance. Flags: --json.
sessions List your work sessions. Flags: --all (every contributor's), --json.
capabilities Report this folder's tracking state and what to do next (for AI agents). Flags: --json.
matrix (alias integrations) Show the integration capability matrix. Flags: --json. (Maintainer-only: --write-readme, --verify-live.)
report Generate a shareable report. Flags: --format <html\|md\|json> (default html), --open, --author <slug>, --team, --title <text>, --json.
verify Run integrity checks on your trail (config, journal, artifact hashes, report).
trace <file> Show every snapshot and related event for a file. Flags: --format <text\|json> (default text).

Manage the inbox

Work Showtail captured but couldn't place in a project (folderless / scratch sessions) waits in the inbox. By default only real-project, signal-bearing work shows; the rest is kept aside (recoverable).

Command What it does
inbox List real-project sessions awaiting placement; pick to place them, or dismiss (d1,3 / dismiss all). Flags: --all (also show scratch kept aside, tagged with why), --json.
ignore [path] Mark a folder as scratch so its sessions never surface in inbox (still under --all). No path lists ignored folders. Flags: --remove, --list, --json.
move [sessionId] (alias reattach) Move any captured session to another project folder. With no id, lists every session to pick from. Flags: --to <path>, --json.

Connect your tools

Command What it does
connect <tool> Connect an AI tool so your prompts and edits are captured (claude, codex, copilot, …). Flags vary by tool: --user, --project, --no-hooks, --extension, --yes, --force. Inapplicable flags are rejected loudly.
disconnect <tool> Remove a tool's instructions/skill and any auto-capture hooks. Flags: --user, --project (default).
import <tool> [source] Import conversations from another tool into your trail. Subcommands below.
import undo Undo the most recent import (permanently removes that batch of events).

import subcommands

Share-based tools (chatgpt, gemini) accept a share URL or a paste:

--no-responses · --paste · --clipboard · -y, --yes · --file <path> · --date <yyyy-mm-dd> · -s, --session <id>

Transcript-based tools (claude / alias claude-code, codex) read an on-disk transcript:

--list · --no-responses · --file <path> · -s, --session <id>

See each tool's integration guide for worked examples.

Exit codes

Code Meaning
0 Success.
1 General error.
2 Not initialized (run showtail track, or just start working with a connected tool).
3 verify found a problem with the trail.

(A ShowtailError may set its own specific code; agents can branch on these.)