From a topic to a cited, human-sounding report

The Research Agent plans the research with you, searches Perplexity, Exa and Tako in parallel, reads the sources, and writes a report whose citations are audited in code — then humanizes and scores it before you see it. Not a bypass tool. A research partner that writes like you.

3 search providers in parallel Citations audited in code Humanized + scored 1–20 pages, English & Arabic
What a run looks like

You approve the plan. The agent does the legwork.

Every stage streams live — nothing happens behind a spinner.

How it works

Four steps, you in control at every hand-off

01

Describe it or upload it

Type a topic or a few questions, or attach your own PDFs and notes — they become citable source material alongside live web results.

02

Approve the outline

You see the plan before anything runs: sections, source scope, quick or deep, and a 1–20 page length target. Edit any of it.

03

Watch the research happen

Perplexity, Exa and Tako are searched in parallel for every query; the top sources are read in full. Planning, reading and writing stream live with a stage tracker.

04

Humanized, scored, delivered

The draft is rewritten with Max or Sage until it scores green, then Writing Pro scores it. Edit by chat, listen, export slides or PDF, share a link.

Why it is different

Research tools research. This one delivers.

Most assistants stop at a list of links. The Research Agent ends with a draft you can submit.

The best citations on the web

Three providers, one query

Perplexity, Exa and Tako searched at once, merged and de-duplicated. Full-text reads for the top sources, not snippets. Your PDFs join the same pool.

Citations that cannot be invented

Audited in code, not prompts

The writer may only cite registered sources. Markers are rebuilt from the delivered text after humanizing, so numbers always match the source list.

Reads like you wrote it

Humanized and scored automatically

Max or Sage rewrite the draft until it reads human; Writing Pro Expert returns the score before you ever see the report.

Delivered in every format

Report · audio · slides · PDF

Evidence table with CSV export, audio overview, slides view with PowerPoint export, PDF and Markdown download, share links, chat edits.

Same engine as the plagiarism checker

One source of truth

The providers behind the Agent also power Humanizily's plagiarism checks, together with OpenAlex, Europe PMC and Wikipedia.

English and Arabic

RTL interface included

Research Arabic topics in an Arabic UI and hand the draft to the Arabic-aware humanizer and the tashkeel / i'rab suite.

Straight talk

What we promise — and what we don't

We promise real sources, citations that exist, and a draft that reads naturally. We do not promise a specific score on any third-party detector — detectors change constantly and nobody can guarantee them. The Writing Pro score tells you honestly where the draft stands; the evidence table lets you check every claim yourself.

Free trial: 2 Research Agent runs and 500 words, no card. Unlimited — $39/month or $250/year — has no run limits. Cancel anytime.

FAQ

Research Agent questions

What does the Research Agent actually do?

You give it a topic, a few questions, or your own PDFs and notes. It proposes an outline you can edit and approve, then researches each section across Perplexity, Exa and Tako at the same time, reads the full text of the best sources, and writes a report with numbered citations. The draft is then humanized with Max or Sage and scored by Writing Pro before you see it.

Can it invent citations?

No. The writer can only cite sources that were actually retrieved and registered during the run; the citation audit is pure code, not a prompt. Citation markers are rebuilt from the delivered text after humanizing, so the numbers on the page always match the source list.

Which search providers does it use?

Perplexity, Exa and Tako run in parallel for every query. Results are merged and de-duplicated, and the top sources are fetched in full text rather than read from snippets. Your own uploaded PDFs and notes are added to the same source pool.

How long is a report and how long does a run take?

You choose quick or deep research and a length target from 1 to 20 pages before the run starts. Most runs finish in a few minutes; the live view shows every stage — planning, searching, reading, writing, humanizing, scoring — with pause, resume and cancel.

What can I do with the finished report?

Edit any section by chat, ask follow-up questions on the sources, open the evidence table (sortable, CSV export), listen to an audio overview, switch to a slides view and export PowerPoint, download PDF or Markdown, or share a link. The report can also open in the Humanizily editor for LaTeX, Word export and plagiarism checking.

Is the Research Agent included in the price?

Yes. Every new account gets 2 free Research Agent runs (plus 500 words, no card). Unlimited — $39 a month or $250 a year — removes the run limit and includes everything else, alongside the humanizers, Writing Pro Expert detector, plagiarism checks, editor and every tool.

Does it work in Arabic?

Yes. The Agent has an Arabic interface with RTL layout, researches Arabic topics, and hands the draft to the Arabic-aware humanizer. The wider Arabic suite (tashkeel, i'rab, sarf, balagha) is available in the same workspace.

Get started

Your next report, researched and written like you.

Included in Humanizily Unlimited with the humanizers, detector, plagiarism checks, editor and 21 tools.

Get Unlimited

Secure Stripe checkout. Cancel anytime.