Manus

Manus is a general-purpose AI agent platform built around execution rather than chat alone. Current Manus 1.6 offers Lite, standard and Max agent variants for multi-step work such as web research, spreadsheet/data analysis, slide creation, website and mobile development, image/design work and other artifact-producing tasks. Tasks can run asynchronously in the cloud while the user leaves the app.¹²

The execution surface now extends beyond Manus's cloud sandbox. Browser Operator can work inside an explicitly authorized local browser session using existing logins and premium tools, while the desktop “My Computer” feature can operate on authorized local files and command-line tools. Manus also exposes an API for persistent custom agents and tasks, including a default general-purpose agent.³⁴⁵

The current Help Center lists Free at $0/month with Chat Mode plus Manus 1.6 Lite in Agent Mode and 300 daily-refresh credits, subject to a 1,500-credit monthly consumption cap for those daily credits. Pro starts at $20/month on annual billing with 4,000 monthly credits; a higher Pro tier starts at $40/month with 8,000 monthly credits and a seven-day trial. Team starts at $20/seat/month on annual billing and adds collaboration, SSO, data-training opt-out, usage analytics and internal access controls.¹⁶

Credit behavior matters. Monthly subscription credits reset rather than roll over, while add-on credits remain usable only while a paid subscription stays active. Complex tasks can consume materially more credits than simple chat, so effective cost depends heavily on workload.⁶

The official iOS app provides a substantial product-specific adoption signal: the India storefront recently showed about 4.7/5 from roughly 2.1k ratings, with other country storefronts showing similarly high but locally different aggregates.⁷ G2's exact Manus AI Agent profile is smaller but still positive at about 4.6/5 from 11 reviews, with praise for research, presentations and asynchronous task execution alongside complaints about price, credits, support and output correction.⁸

Service sentiment is much worse on Trustpilot. The exact manus.im profile is around 1.2/5 from roughly 196 reviews, with recent complaints concentrated on billing, cancellations, support and credit consumption.⁹ That signal is useful for service risk, but it is not interchangeable with App Store or G2 product ratings and should not be converted directly into the same numeric product score.

Independent hands-on evidence is mixed. VentureBeat's 2025 data-visualization test found strong messy-data handling but weak transformation lineage for enterprise use, while TechRadar found Manus effective at building integrated creative packages but slower and imperfect on some image details.¹⁰¹¹ Those tests predate Manus 1.6 and are therefore historical context rather than current benchmarks. Manus's own 1.6 Max performance charts are first-party internal benchmarks and are not treated as independent proof.²

Corporate status is also unusually fluid. Reuters reported on August 11, 2026 that Manus would resume independent operations as its Meta deal unwound, and the Financial Times subsequently reported a buyback structure in which Tencent was expected to become the largest minority shareholder while Manus remained operationally independent.¹²¹³ Some Manus/App Store surfaces still carry Meta branding, so ownership/branding metadata should be treated as transitional rather than perfectly synchronized. The Verge's reporting on undisclosed creator promotions and aggressive earnings messaging adds a separate commercial-trust concern.¹⁴

Manus is best for users who want to delegate complete multi-step deliverables rather than request individual answers: broad research, data work, reports, slide decks, web/mobile prototypes, design/media tasks and browser workflows. The mobile, browser and desktop surfaces make it unusually flexible across environments.

It is a weaker fit where predictable fixed-cost usage, mature human support, strict transformation audit trails or highly controlled autonomous actions are mandatory. High-stakes work should still be reviewed before relying on the result, especially when Manus has acted on external systems or transformed source data.