Quickstart¶
The whole flow is connect once, then work. After a one-time setup, your prompts and edits are captured automatically and you finish by generating a report.
showtail setup # one-time: connect your AI tools and turn on automatic tracking
# ...just work — your prompts and the files your tool edits are captured for you...
showtail report # generate the report for your educator
showtail setup is the fastest start: it connects the AI tools it finds and
turns on automatic tracking for every project. Once a tool is connected,
Showtail initializes a project for you the first time you work in it — the
.showtail/ folder is created on the spot, so there is no separate init step to
remember.
Wiring up one project by hand¶
Prefer to set up a single project explicitly instead of turning on tracking
everywhere? Use track and connect:
# 1. Set up Showtail in your project. This creates the .showtail/ folder.
showtail track --project "Week 5 Parser"
# 2. Connect your AI tool so capture is automatic (one time).
showtail connect claude # or: codex, copilot
showtail track is the manual path. If you already ran showtail setup,
tracking is on everywhere and you can skip both steps.
While you work¶
Your prompts and the files your tool edits are captured for you. Check where you are at any time:
showtail status # current session, event count, connected tools
showtail sessions # list your work sessions (--all for the whole team)
showtail trace src/parser.ts # the full trail for one file
Generate a report¶
showtail report # HTML + Markdown in .showtail/reports/ (team + per-student)
showtail report --format md # Markdown only
showtail report --format json # Machine-readable JSON
showtail report --team # just the combined team report
showtail report --author <slug> # just one student's report
Before you submit¶
showtail verify # integrity checks on your trail
showtail end # close the current session
Commit .showtail/ with your work so your educator can review it.
Not using a tool with hooks?
ChatGPT and Gemini can't run commands on your machine, so they are import-based — you bring a shared conversation into the trail after the fact. See ChatGPT and Google Gemini. You can also back-fill an earlier Claude Code session.
Next steps¶
- Integrations — per-tool setup and the capability matrix.
- How it works — the event model behind the trail.
- For educators — running this in a class.
- CLI reference — every command and flag.