An autonomous AI agent platform for complex, multi-step task execution across research, development, automation, and analysis.
<ul><li><strong>Pricing Model:</strong> Freemium with variable-rate credit consumption (driven by LLM tokens, compute, and third-party APIs) + separate usage-based website hosting with free monthly allowances</li><li><strong>Packaging Model:</strong> Free / Pro (three volume sub-tiers: $20/mo for 4K credits, $40/mo for 8K, $200/mo for 40K) / Team ($20/seat/mo, 2-member min, shared credit pools) / Enterprise (custom)</li><li><strong>Credit Model:</strong> Monthly credit pool (no rollover) + 300 daily refresh credits (expire in 24hrs) + add-on credit packs (non-expiring, require active subscription); Team plan uses shared pooled credits across seats</li></ul>
September 16, 2026
Last update:
<h3>Product Overview</h3><p>Manus is an autonomous AI agent platform, now operating under Meta ownership following a ~$2.5 billion acquisition in December 2025, that executes complex, multi-step tasks end-to-end without requiring continuous human supervision. Unlike AI chatbots that respond to individual prompts, Manus operates inside isolated cloud sandboxes to plan, execute, and deliver complete work products: full-stack web applications, research reports, slide decks, browser automation workflows, and data analysis outputs. The platform is built on top of Anthropic's Claude 3.7 Sonnet and customized versions of Alibaba's Qwen, rather than a proprietary LLM, and has processed over 147 trillion tokens across 80 million virtual computers since launch.</p>
<h3>Key Features & Capabilities</h3><p>Manus reached $100 million in ARR within approximately eight months of its first paid plan launch (a claimed record pace), growing at over 20% month-over-month since the release of Manus 1.5. The platform is listed among the MIT AI Index's top 30 AI agents alongside Claude, ChatGPT, Gemini, and Salesforce Agentforce, and has outperformed OpenAI's Deep Research (o3) by more than 10% on the GAIA benchmark for real-world multi-step task evaluation. Post-acquisition, Manus is continuing to operate its standalone B2B subscription while integrating into Meta's broader consumer and business product ecosystem. These capabilities span agent execution, development workflows, browser automation, integrations, and enterprise controls, with plan access differentiated by model tier, credit availability, and team administration features.</p><ul><li>Core agent execution and model tiers: Plans, executes, and delivers complete work products end-to-end without continuous human oversight; operates in isolated cloud sandboxes with no software installation required. Wide Research spins up multiple parallel sub-agents simultaneously to conduct large-scale web research; all Wide Research sub-agents run on Manus 1.6 Max; available on all paid tiers. Model access spans manus-1.6-lite for simple tasks and daily refresh usage, manus-1.6 for general tasks, and manus-1.6-max for complex analysis. Daily refresh credits (300/day) can only be used for Manus 1.6 Lite tasks; tasks using manus-1.6 or manus-1.6-max draw exclusively from monthly, add-on, or free credits.</li><li>Development, content, and data workflows: Supports full-stack web development with backend server infrastructure, user authentication, persistent databases, embedded AI, custom domain support, built-in analytics, mobile app development, and visual point-and-click editing on live preview. It also supports slides (PPT/PDF), documents (Word/Markdown/PDF), spreadsheets (Excel/CSV), CSV analysis, data cleaning, charts and visualization generation, and payment configuration through Stripe products, pricing, and webhook configuration via Claimable Sandbox.</li><li>Automation, integrations, and browser operation: Browser Operator works within the user's local browser session using existing logins (not a remote cloud browser) and executes from the user's local IP address, enabling access to platforms with strict IP-based security controls. It supports Chrome and Edge and is available to all users at no additional charge. Supported workflow types include form filling, multi-site navigation, data extraction, scheduled tasks, email automation, and Slack task execution. Integrations include Gmail, Google Calendar, Google Drive, Notion, Slack, Stripe, Zapier, Meta Ads Manager, Instagram, Instagram Creator Marketplace, Telegram, MCP Connectors, and Similarweb.</li><li>Enterprise controls and security: Certifications include SOC 2 Type I & II, ISO 27001, and ISO 27701. Model providers are prohibited from training on Team/Enterprise data, with no model training on user data. Subprocessors include Anthropic, Google Cloud Platform, Microsoft Azure AI Foundry, and Amazon Web Services. Admin controls include unified billing management, usage dashboards, and granular sharing controls; SSO is available as a Team plan add-on; a startup program is available at manus.im/startups.</li></ul>
<h3>Pricing Model Analysis</h3><p>Manus uses an action-based credit consumption model layered on a Good/Better/Best subscription structure. Customers purchase a tier that includes a monthly credit pool; each AI agent task consumes credits based on computational complexity: LLM tokens, virtual machine time, and third-party API calls. A separate usage-based billing layer applies to website builder hosting, creating a dual-billing architecture that requires TCO modeling for teams deploying Manus-built applications at scale.</p><div class="tableResponsive"><table cellpadding="6" cellspacing="0"><tr><th>Metric Type</th><th>What Measured</th><th>Why It Matters</th></tr><tr><td>Value Metric</td><td>Credits consumed per task</td><td>Directly ties cost to work performed; complex tasks (e.g., web apps) consume 6-10x more credits than simple tasks</td></tr><tr><td>Usage Metric</td><td>Monthly credits + daily refresh credits (300/day)</td><td>Monthly pool drives tier selection; daily refresh is a rate-limiting mechanic uniform across all tiers</td></tr><tr><td>Billable Metric</td><td>LLM tokens + VM compute time + third-party API calls</td><td>Three distinct cost drivers bundled into a single credit unit; per-task costs are not individually visible</td></tr></table></div>
<h3>Customer Sentiment Highlights</h3><ul><li>“I'm going through 200k++ credits in Manus every month and it's worth every penny.”<b> <span class="pricingHiphenSymb"> - </span>Twitter/X User</b></li><li>“Because it's cheap and saves me a bunch of time. Cheaper than a junior team member and delivers similar results for many tasks.”<b> <span class="pricingHiphenSymb"> - </span>Reddit User</b></li><li>“I'm genuinely impressed by its high efficiency and task completion, it can quickly and accurately build product prototypes based on my needs.”<b> <span class="pricingHiphenSymb"> - </span>Product Hunt Reviewer</b></li><li>“transformed how i think of myself as a builder and enabled so much internal tooling that has been a joy to build and use”<b> <span class="pricingHiphenSymb"> - </span>Product Hunt Reviewer</b></li><li>“The price is great $39 a month. Well worth the cost for what it's capable of doing.”<b> <span class="pricingHiphenSymb"> - </span>Reddit User</b></li></ul>
Metronome’s Take
<p>Manus operates a credit-based subscription model where every agent action — browsing, code execution, data analysis, file operations — consumes credits at a variable rate driven by LLM token usage, virtual machine compute, and third-party API calls. The platform offers three plan levels: Free (300 daily credits, no monthly pool), Pro (three volume sub-tiers from $20/month for 4,000 credits to $200/month for 40,000), and Team ($20/seat/month with shared credit pools and SSO). All plans include 300 daily refresh credits that reset every 24 hours, separate from the monthly pool. Website hosting operates on a distinct usage-based billing track with free monthly allowances that cover most projects until they hit significant traffic. Monthly credits do not roll over; only purchased add-on packs carry forward indefinitely.</p>
<p><strong>Recommendation:</strong> The core tension in Manus's model is that it prices on computational complexity in an environment where that complexity is opaque to the buyer at the point of consumption. This works when users develop intuition for task cost over time, but it creates onboarding friction and budget unpredictability that flat-rate or fixed-per-unit models avoid. Companies considering a similar variable-rate credit model should invest heavily in pre-task cost estimation and real-time consumption visibility — the absence of both in Manus is the most consistent criticism in user reviews. The separate website hosting billing track is also worth noting as a design pattern: splitting creation (credits) from distribution (usage-based hosting) lets Manus monetize two distinct value moments without overloading a single metering dimension, but it introduces a second cost surface that buyers need to model.</p>
<h4>Key Insights</h4><ul><li> <strong>Variable-rate credit consumption tied to computational complexity, not fixed per-task pricing.:</strong> A simple chat lookup costs 10–50 credits while a deep autonomous research task can burn 500–1,000 credits, with an approximate rate of 11–14 credits per minute depending on complexity. Critically, Manus does not surface pre-task cost estimates or real-time consumption alerts — users learn actual credit spend only after task completion. This creates a fundamentally different cost predictability profile than platforms with fixed per-unit pricing (like Suno's 5 credits/song), and means budgeting requires historical usage patterns rather than upfront calculation.</li><li> <strong>Three-layer credit architecture separates daily, monthly, and purchased credits.:</strong> Daily refresh credits (300/day) provide a baseline floor that resets every 24 hours even if the monthly pool is exhausted; monthly pool credits expire at cycle end with no rollover; and purchased add-on credits persist indefinitely but require an active subscription. This layered design gives users a guaranteed daily minimum while creating clear upgrade triggers when daily credits prove insufficient — but the three distinct expiration policies add real tracking complexity.</li><li> <strong>Team plan introduces shared credit pooling with per-seat pricing.:</strong> The Team tier ($20/seat/month, 2-member minimum) pools credits across members rather than siloing them per user, with centralized billing, SSO, usage analytics, and admin visibility into who is consuming credits on what. For teams with uneven workload distribution — where one member runs heavy research while others do lighter tasks — shared pooling is materially more efficient than per-seat allocation, though it also means one power user can drain the team's pool without per-user guardrails.</li></ul>
The Pricing Experimentation Playbook
Find your ideal pricing model
Answer 8 quick questions to discover which best fits how your customers get value from your product.
