FlashData CLI

简体中文 · API docs · Pricing

Google Search and YouTube data for your terminal and AI agent. Search the web, retrieve video metadata and transcripts, or process video lists with bulk jobs.

Install with uv and set up your agent:

uv tool install flashdata-cli
flashdata setup

Choose Codex, Claude Code or Cursor and confirm the project directory. Setup installs the skills and configures access without a paid query.

Try before signing up. Eligible new users can start with temporary access and limited Credits for Google Search and YouTube metadata. Existing credentials are reused. Already have a key? Run flashdata auth login before setup. See account setup for trial limits and signing up later.

Run your first search:

flashdata search "battery recycling startups" --num 5

For terminal-only use, run flashdata init --name "My CLI" instead of setup. Data queries use Credits; flashdata account balance shows your remaining allowance.

python -m pip install --upgrade flashdata-cli

uv pip install --upgrade flashdata-cli

brew install flashdatadev/tap/flashdata

Save Google search results:

flashdata search "battery recycling startups" --num 5 --output search.json

Find YouTube videos, then use a returned video ID to get metadata or an existing transcript:

flashdata youtube search "Go programming tutorial" --max-results 5
flashdata youtube metadata dQw4w9WgXcQ --output video.json
flashdata youtube transcript dQw4w9WgXcQ --language en --output transcript.json

Replace the example video ID with your own. Transcript availability depends on the video's language tracks. Use youtube captions for uploader captions.

For Google news, images, shopping, videos, places and other YouTube sources, run flashdata sources list, then flashdata sources show SOURCE. Submit the documented input with flashdata query --input query.json. See the command reference and request examples.

CLI queries return JSON by default. Choose plain text or Markdown when needed:

flashdata search "vector databases" --accept text/plain
flashdata search "vector databases" --accept text/plain --respond-with markdown

--output FILE saves the selected response body to a new file. --json wraps text in the CLI's JSON envelope. Agent skills request plain text by default; bulk exports always use JSONL. See response formats for async jobs and batches.

Create ids.txt with one YouTube video ID per line:

dQw4w9WgXcQ
jNQXAC9IVRw

Export metadata for the list:

flashdata bulk run --source youtube_metadata --input ids.txt --run-id video-dataset-001 --timeout 1h --output videos.jsonl

The command submits the job, waits and saves one JSONL row per video. Bulk jobs also support transcripts, captions, license classifications and downloads. Downloads deliver media to your cloud storage; the JSONL export contains delivery receipts.

If interrupted, recover the same job into a new file:

flashdata runs recover video-dataset-001 --timeout 1h --output recovered-videos.jsonl

Recovery reads existing work without submitting another job. See the bulk guide for MP3/MP4 examples, storage setup, failed items and job controls.

uv tool upgrade flashdata-cli
flashdata setup

Setup refreshes unmodified skills and preserves local edits. For other installers, follow the upgrade guide.

• Agent setup: select skills, install globally, or use --skip-auth to install skills offline without changing credentials. MCP is optional.
• Account setup: temporary access, API keys and connecting your account.
• Request examples: background queries, batches and full source options.
• Credentials and automation · Run recovery · Contributing

Report an issue · Releases · MIT license