Strip page and JSON noise before the Bot reads it

How this helps you

You spend weekly tokens on the answer, not on ads, nav, and unused JSON.

When I give you a URL, call the page-cleaner tool and answer only from the cleaned article text. When I give you JSON or an API payload, call the field picker and keep only these fields: [fields]. Summarise what you dropped. If the cleaned input is missing something critical, say what is missing instead of guessing.

What they did

In a long usage-saving video, Jordan's first demos are Trafilatura-style page cleaning (Wikipedia and product pages reduced to article or product words) and jq-style field picking from API or repo JSON. The Bot still does the reasoning; the tool drops buttons, ads, images, and unused fields before tokens are spent. His headline 98.7% cut is from his own paired tests on those tasks, not a promise for every workflow. Commenters separately beg for one-click archive of huge Bot chats, which pairs well with this hygiene. Install once, then keep the standing rule so every future URL job stays cheap.

Who it's for: Power users who paste big pages or API dumps into Bots and burn the week early.

Before you start

• Ability to install or connect local CLI tools the Bot can call
• A repeat job that currently pastes whole web pages or big JSON
• Willingness to verify the cleaned output still answers the question

How to try it

1. Pick one noisy input you repeat (web article, docs page, or API JSON).
2. Install a cleaner for pages and a field picker for JSON on the machine the Bot can reach.
3. Teach the Bot: for URL questions, call the cleaner and reason only on the cleaned text; for JSON, request named fields only.
4. Run the same question with and without cleaning once, and compare usage and answer quality.
5. Add a standing rule: never paste raw HTML or full API dumps into chat when a tool can narrow them.

What to expect

You ask about a long docs page; the Bot cleans it first and answers from the article body instead of re-reading navigation chrome.

Common mistakes

• Treating a creator's percentage claim as your guaranteed weekly saving.
• Cleaning so hard that citations and tables disappear.
• Ignoring fat Bot chat history while only trimming page inputs.

Related tips

• A read-only usage audit across your whole Bot team: One prompt has your chief of staff check every Bot's routines, schedules and triggers, then report what to pause, slow, narrow or merge, and change nothing until you approve.
• Analyse your own usage CSV: Export usage from the Usage page and ask your Chief of Staff Bot how to cut it.
• A /clean-start skill for long chats: Save the goal, decisions, files and next action to a handoff file, then pick up in a fresh chat.

Worth knowingYou spend weekly tokens on the answer, not on ads, nav, and unused JSON.

Learn more: Grok Bot Use Cases: Real Workflows From Real Users · All Grok Bot workflows tips