The job
You’re staring at a browser full of sources and you need a usable bibliography now, not after twenty minutes of manual cleanup. You want Sephir to read your open tabs, pull the citation fields that matter, and return a clean list you can paste into your draft.
Why this is hard without Sephir
Without Sephir, you bounce tab to tab copying titles, authors, dates, and links into a separate tool, then fix broken fields by hand. A regular chatbot can’t see your open tabs, so you end up pasting partial text chunks and losing source boundaries. By the fourth or fifth source, the thread turns noisy and citation formatting errors creep in.
How Sephir does it
- Pin the tabs you want cited so the agent has a fixed source set.
- Open the Sephir sidepanel with
Cmd+Shift+S. - Ask for your target format (APA, MLA, Chicago) across all pinned tabs.
- Watch the timeline run
extractPageText(tab)on each source and merge the results. - Review the draft bibliography and ask for strict field checks before finalizing.
- Export the final list to Markdown or JSON, then save the trace as
/cite-tabs.
The skill behind it
This skill runs a repeatable extraction flow across pinned sources, then formats results into a bibliography you can reuse. It keeps the process explicit and auditable, not a black box.
What it costs
Sephir runs this recipe on your own ChatGPT Plus via Codex OAuth or your own API key, so inference billing stays with your provider. Typical token usage: ~3,000-8,000 input + ~500-1,500 output per run on Claude Opus 4, GPT-5, or Gemini 3 Pro. Check plan details on ; this workflow is best with Pro Lifetime features.