Consensus

Consensus isn't strictly autonomous in the agent sense - it doesn't iterate or plan multi-step workflows. It's a search engine with a clever trick: it scans academic papers and gives you a consensus meter showing whether studies agree, disagree, or are split on a question. You ask a plain-English question, it returns a summary plus a percentage breakdown. The autonomy is minimal - one query, one answer - but the value is in aggregation speed. I've used it mostly for health claims that show up in newsletters or podcasts. Someone says "intermittent fasting improves insulin sensitivity" and I want to know if that's backed by actual trials or just one cherry-picked study. Consensus returns a grid of papers with yes/no/maybe classifications and a consensus score. It's faster than PubMed because it pre-digests the abstracts and highlights the relevant finding. The summaries are usually accurate, though I spot-check the original papers when stakes are high. Where it shines: quick sanity checks before you write something, or when you need to deflate a confident claim from someone who clearly hasn't read the literature. It's also useful for scoping a topic before a deeper dive - if consensus is 80% one way, you know where the weight of evidence sits. The free tier gives you a handful of searches; the paid version unlocks unlimited queries and better filtering. Failure modes: it's only as good as the papers indexed, and coverage outside biomedicine and social science is patchy. I've asked niche engineering questions and got thin results. The consensus meter can also mislead if sample size is small - three papers agreeing doesn't mean much if there are thirty others it missed. And it won't catch nuance like "works in mice, fails in humans" unless you drill into individual studies. Nearest competitor is probably Elicit, which does similar literature synthesis but leans harder into extracting structured data from papers. Consensus is faster for yes/no questions; Elicit is better when you need tables or comparisons across studies. I'd pick Consensus for quick fact-checks, Elicit for actual research projects.