⠿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣶⣄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⣠⣶⣿⣿⣿⣿⣿⣷⣦⡀

⠀⠹⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣆⠀⠀⠀⠀⠀⠀⠀⠀⣰⣿⣿⣿⣿⣿⣿⣿⣿⣿⣷

⠀⠀⠙⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣆⠀⠀⠀⠀⠀⠀⣰⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡟

⠀⠀⠀⠀⢻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣆⠀⠀⠀⠀⣰⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡟⠀

⠀⠀⠀⠀⠀⢻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡆⠀⣠⣾⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠟⠀⠀

⠀⠀⠀⠀⠀⠀⢻⣿⣿⣿⣿⣿⣿⣿⣿⣿⣷⣶⣿⣿⣿⣿⣿⣿⣿⣿⣿⡿⠟⠁⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠙⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡟⠁⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠀⠈⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⡄⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣦⣀⠀⠀⠀⠀⠀⠀⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⢿⣿⣿⣿⣿⣿⣿⣿⣿⣿⣿⠛⠀⠀⠀⠀⠀⠀⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠈⠻⣿⣿⣿⣿⣿⣿⣿⣿⠃⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠘⣿⣿⣿⣿⣿⣿⠃⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠘⣿⣿⣿⡿⠃⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀

⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⢸⣿⠋⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀⠀

Virlo CLI — run your social intelligence from the terminal.

The official command-line interface for Virlo: TikTok, YouTube Shorts, Instagram Reels, and Meta Ads research — trends, viral videos, creators, sounds, hashtags — designed to be driven by humans (interactive setup wizard) and by scripts / AI agents (subcommands, flags, --json).

Install

npm install -g virlo-cli

Or run it without installing anything:

npx virlo-cli <command>

Requires Node ≥ 20.

Quick start

virlo setup             # interactive wizard: paste your virlo_tkn_… key, it validates and saves
virlo whoami            # masked key + live balance
virlo trends digest     # today's trends 💲 $0.25 (prompts before spending)

Kick off a keyword search, wait for it, then read the results:

virlo orbit create \
  --name "ai tools" \
  --keywords "ai tools,ai apps" \
  --time-period this_week \
  --platforms tiktok,youtube \
  --yes --watch

virlo orbit videos <orbit_id> --order-by views --sort desc --limit 20
virlo orbit analysis <orbit_id> --latest

No key yet? Sign up at dev.virlo.ai, then Dashboard → API Keys → Generate Key.

What you can do

• Spot trends — virlo trends digest, virlo videos digest 💲 $0.25 · virlo hashtags list 💲 $0.05
• Research a niche — virlo orbit create 💲 one-shot keyword search across platforms
• Keep watching a niche — virlo comet create 💲 recurring scheduled scrapes
• Deep-dive a creator — virlo satellite creator tiktok @handle --include videos,outliers 💲
• Explain a viral video — virlo satellite video-outlier <url> --platform tiktok 💲
• Research sounds — virlo sounds trending / search / usage-history 💲
• Track creators & videos over time — virlo tracking creators add, virlo tracking videos add 💲
• Get notified — virlo webhooks create --url … --events comet.run.completed,…
• Hand it to an agent — virlo skill install teaches Claude the whole surface

Usage

Authentication & configuration

Effective config is resolved per-call as flag > env var > file > default:

Setting	Flag	Env var	File key	Default
API key	--api-key	VIRLO_API_KEY	apiKey	(none — required)
Base URL	--base-url	VIRLO_BASE_URL	baseUrl	https://api.virlo.ai

virlo setup validates your key against the free balance endpoint and saves it to ~/.virlo/config.json. virlo whoami shows what's configured (and where it came from); virlo logout deletes it.

VIRLO_CONFIG_DIR overrides the config directory (defaults to ~/.virlo). VIRLO_TIMEOUT_MS overrides the per-request timeout (default 60000).

Security note: the API key is stored in plaintext at ~/.virlo/config.json with chmod 600 (owner read/write only). Anyone who can read that file can spend your credits. For CI or shared machines, prefer the VIRLO_API_KEY env var (nothing is written to disk), and rotate keys from the dashboard if exposed.

Global flags

Flag	Description
--json	pure JSON on stdout — for scripting / agents
-y, --yes	confirm paid commands without prompting (required when non-interactive/--json)
--api-key	override the configured API key
--base-url	override the API base URL
-v, --verbose	log requests to stderr (API key redacted)

Global flags parse before or after the command: virlo --json orbit list ≡ virlo orbit list --json.

Output & scripting

• Default: human-readable tables / key-value blocks on stdout.
• **--json**: the response data payload as pure JSON on stdout (nothing else), so it pipes cleanly into jq or an agent. Errors go to stderr as {"error":{"type","message",…}} — including usage errors. When a list response paginates, one {"pagination":{…}} line goes to stderr (check has_next_page/ total there); spend notices also go to stderr, keeping stdout machine-clean.

Exit codes, for branching in scripts:

Code	Meaning
0	ok
1	generic error
2	usage / validation
3	auth / config
4	insufficient credits
5	rate limited
6	not found
7	network

Money & spend safety

💲 commands require confirmation before they spend. Running one interactively in a terminal prompts a y/N confirm (default No). Non-interactive callers — including Claude and scripts — and any call using --json must pass -y/--yes up front; otherwise the command fails with a validation error before any request is made, no credits spent. All paid commands route through src/lib/spend-guard.ts, the single place this gating lives (per-session/daily spend caps and --dry-run are still future work).

Costs are resolved server-side and may change; known cost hints are shown in --help and printed before paid calls, and the actual charge is always reported from the response headers (💲 spent $X.XX · $Y.YY remaining, on stderr). Existing results are free to re-read — check virlo orbit list, virlo comet list, virlo tracking creators list before creating new paid work.

Async jobs (--watch)

orbit create, satellite lookups, and tracking collect are asynchronous: the create call returns an id immediately. Add --watch to poll until the job reaches a terminal state (transient network blips during the wait are retried). Note that orbit/comet report completed when the scrape finishes — the AI analysis may still be generating (fields come back null); re-check with virlo orbit analysis <id> --latest.

Agent skill (for Claude / AI agents)

Install a SKILL.md that gives an AI agent (Claude Code) full context on Virlo and this CLI — capabilities, the cost/safety model, the async model, interpretation rules, and intent routing — so it can drive virlo correctly.

virlo skill install            # interactive: pick project / global / custom dir
virlo skill install --global   # ~/.claude/skills/virlo/SKILL.md (all projects)
virlo skill install --dir ./.claude/skills/virlo   # explicit, non-interactive
virlo skill print              # dump the skill markdown to stdout

Place it under .claude/skills/<name>/SKILL.md for Claude Code to auto-discover it. Add --force to overwrite an existing file.

Command reference

Run virlo <group> --help for full options. 💲 = spends credits.

Group	Highlights
account	balance
hashtags	💲list, 💲performance <hashtag>
videos	💲digest
trends	💲list, 💲digest
orbit	💲create, list, get, videos/slideshows/ads/outliers/sounds/trends/analysis <id>
comet	💲create, list, get, update, delete, + same sub-resources as orbit
satellite	💲creator <platform> <user>, creator-status, 💲batch, batch-status, 💲video-outlier <url>, video-outlier-status
sounds	💲trending, 💲search <q>, 💲by-creator, 💲get, 💲usage-history, 💲videos
tracking	creators / videos sub-groups: 💲add, list, get, report, snapshots, posts, 💲collect, update, remove
webhooks	create, list, get, update, delete, reenable, test, deliveries, retry-delivery

Out of scope (different auth, not a normal API key): /v1/admin/* and /v1/teams/*.

Development

git clone https://github.com/Virlo-AI/virlo-cli && cd virlo-cli
npm install
npm run build
npm link            # makes your local build available globally as `virlo`
npm run virlo -- …  # or run from source without building (tsx)

src/
  index.ts                 commander root + global flags + exit-code mapping
  config/                  ~/.virlo config (chmod 600) + resolution
  client/                  fetch wrapper, envelope unwrap, error mapping, timeout
  lib/                     output renderer, flags/parsers, polling, spend-guard, validation
  domain/endpoints.ts      paid-endpoint registry (cost hints / 💲 markers)
  commands/                one module per resource group
test/                      vitest unit + E2E exit-code tests (all offline)

npm run typecheck   # tsc --noEmit
npm test            # builds, then runs the vitest suite
npm run dev         # tsup --watch

Releases are tag-triggered from CI — see RELEASING.md.

License

MIT

---

Commands marked 💲 spend real credits (1 credit = $0.01); re-reading existing results is free — see Money & spend safety.