Manus vs Elicit (2026)

Manus and Elicit both plan and execute multi-step research work with little hand-holding, but they point at completely different targets. Manus is a general-purpose autonomous agent from Butterfly Effect (the team behind the Monica assistant): give it a goal in plain language and it spins up its own cloud computer, browses the live web with a browser operator, writes and runs code, and hands back a finished deliverable: a report, a slide deck, or even a small deployable web app. Its "Wide Research" mode fans a broad question out across many parallel sub-agents to cover more ground than a single pass. Elicit does not touch the open web at all. It is a hosted product built specifically for scientific and academic literature: its Research Agent plans sub-questions and searches over 138 million papers, and a dedicated Systematic Review workflow automates the screening step of a formal literature review (up to 5,000 papers on its Pro tier, 40,000 on Enterprise) against inclusion criteria you set, which Elicit's own pricing page describes as "PRISMA-grade accuracy" at the top tier.

Rigor and reliability diverge just as sharply as scope. Every claim Elicit generates carries a sentence-level citation back to the source paper, and its own documentation is upfront that hallucination remains a real risk despite that grounding: it tells users to verify important findings against the original paper rather than trust a summary alone. Manus's citations are only as good as whatever its browser operator finds on the open web, and independent testing has found a sharper problem: the agent can degrade on long, branching workflows and has produced fabricated "verification" output that looked real but described a state that never existed. Manus is best suited to straightforward, mostly read-only jobs (a research summary, a data pull), not mission-critical or write-heavy work; Elicit is built for a narrower job (published literature) but does that job with an auditable, structured process a general web agent has no equivalent of.

Pricing and stability tell the rest of the story. Manus runs on a credit system: a free tier gives a small daily refill (about 300 credits, roughly one task a day), and paid Pro tiers start around $20/month for larger monthly credit pools, but usage is unpredictable, since a single complex job can burn hundreds to well over a thousand credits. Elicit's Basic tier is free for unlimited paper search and chat, with Research Agent and Systematic Review usage capped until Pro ($49/month) or Scale ($169/month); Enterprise is custom. The two companies' footing also differs: Elicit is a public-benefit corporation that has raised $31M total ($9M seed, $22M Series A) with a straightforward ownership structure, while Manus's backing has been genuinely unsettled: Meta closed a roughly $2B acquisition of Butterfly Effect in December 2025, Chinese regulators ordered it unwound in April 2026, and a Tencent-led buyback was still being negotiated as of July 2026 with no confirmed close date.