30 seconds to install.
Every agent. Every shell.
The CLI is a plain binary that ships as tl-cli on PyPI. Install it however you'd install any other Python tool. Auth it with Google SSO or password. You're done.
Pick your path.
pipx
The one-liner. Auto-updates. Tokens stored in your OS keychain. Works on macOS, Linux, and Windows without touching your system Python.
$ brew install pipx # or: apt install pipx $ pipx install tl-cli $ tl auth login # browser opens, Google SSO or password $ tl describe ✓ you're ready
Plugin install
Installs the CLI plus the tl-data-analyst skill, the tl-analyst agent, and six slash commands into ~/.claude/plugins/tl-cli/. Zero further config — just open Claude Code and start talking.
$ pipx install tl-cli $ tl auth login $ tl setup claude ✓ /tl, /tl-deals, /tl-channels, /tl-brands, /tl-reports, /tl-balance installed
pip or uv
Install into a virtualenv, a CI runner, a Dockerfile, or an Airflow image. TL_API_KEY env var works as a fallback for unattended runs.
$ pip install tl-cli # or, faster: $ uv tool install tl-cli
TL Slack app
OpenClaw, TL's own AI operator, runs this CLI inside our internal Slack today. A public Slack app you can install in your own workspace is next — join the waitlist.
Join the Slack app waitlist →Works with every agent harness.
Auth, without the pain.
The CLI uses PKCE OAuth via Auth0 + CloudFlare. No client secrets stored on your machine. Tokens live in your OS keychain via the keyring Python package — same place your browser stores its auto-fill credentials.
Two login methods: Google SSO (the default) or password. We added password login specifically for academic researchers at HBS who don't use Google accounts — if it works for a Harvard professor, it'll work for you.
TL_API_KEY env var works as a fallback for CI, scripts, or Docker images where interactive login isn't practical.
$ tl auth login # browser opens, you click, you're done ✓ Logged in as alex@acme.com $ tl doctor ✓ version: 0.1.0 ✓ auth: valid ✓ api: reachable (34ms) ✓ credit: 500 available $ tl describe # complete schema, credit table, filters