Comparison
Syniaps or Manus?
The best-known consumer generalist agent, $125m of annualised revenue, bought by Meta then made independent again on Beijing's orders
What Manus sells: "Less structure, more intelligence": you give an objective, Manus plans, browses, writes code, produces a deliverable (a report, a website, slides, an app) and hands it back, billed in credits according to complexity.
Syniaps
4.0 / 5
Manus
2.9 / 5
Surveyed on 15 September 2026 from the publisher's public pages, cited at the foot of this page. Offers change: check with the publisher before you decide.
Manus at a glance
- Who
- Butterfly Effect, Singapore, formerly Beijing, 4 years old
- Founded in 2022 in Beijing by Xiao Hong (with Ji "Peak" Yichao and Zhang Tao), relocated to Singapore in mid-2025, around 100 employees, $85m raised (Benchmark, Tencent, HongShan, ZhenFund) and $125m of annualised revenue at the end of 2025. Bought by Meta for $2bn on 29 December 2025, ordered unwound by the Chinese regulator (NDRC) on 27 April 2026, formally cancelled on 11 June 2026 and back to independence on 1 September 2026; certified SOC 2 Type II, ISO 27001 and ISO 27701.
- What
- A cloud task agent, credits, deliverables
- Each task opens a Linux virtual machine with a browser, a file system and code execution; Wide Research launches dozens of sub-agents in parallel; typical outputs are research, slides (PowerPoint), deployed websites and apps, design, mobile, music and voice (ElevenLabs). Around that: Scheduled Tasks, Projects (a master instruction plus a file base), Skills (SKILL.md), MCP connectors, the Browser Operator extension (local browser), a desktop app (macOS, Windows), a mobile app, an API, Slack and Telegram.
- How it gets set up
- A free account, then you describe each task
- Immediate self-serve signup (email or Google), 1,000 credits given then 300 a day, and you type your first task. Everything else is on the user: write the instruction for each scheduled task with its time and destination (email, Slack, Drive), create a Project with its instructions and upload your files into it, capture a Skill after a successful task, and connect the OAuth connectors one by one. Nothing is proposed spontaneously: the agent waits to be launched.
- Which models
- Claude in front, Qwen in support, no choice
- Manus trains no foundation model: reasoning rests on Anthropic's Claude family, fine-tuned Qwen models (Alibaba) serve the auxiliary calls, and the user only picks the agent level (1.6 Lite, 1.6, 1.6 Max) without ever seeing the model or its version. Inference therefore goes to the United States at Anthropic (and to Alibaba for Qwen), on top of a SaaS whose data is stored in the United States and Singapore.
- Price
- $0 to $200 a month, credits per task
- Free (300 credits a day, one task at a time, the Lite agent), Pro at $20 (4,000 credits a month), $40 (8,000), $200 (40,000), 20 concurrent tasks, 17% off annually, Team at $20 per seat per month with a shared credit pool, and SSO billed separately below 30 seats. A task costs anywhere from a few credits to 500-1,500 for wide research or a website, monthly credits do not roll over, and they are consumed even when the task fails.
- Who it is for
- Freelancers, growth, researchers, prototypers
- Strong among solo operators and small marketing, research and product teams who want a deliverable within the hour (market research, a deck, an MVP, monitoring) with nothing to install. The 2026 reviews advise against using it as a production system for a company: instability, unpredictable credits, absent support.
Point by point
| Criterion | Manus | Syniaps | Advantage |
|---|---|---|---|
| Intelligence | The Claude family imposed, Qwen in support, three agent levels (Lite, 1.6, Max), and an invisible model | Claude Sonnet 5, Opus 5, Fable 5, GPT through Codex or the model of your choice, with Standard, In-depth and Maximum modes | Syniaps |
| "Execute" | A cloud VM per task (browser, code, files), parallel sub-agents, deployed websites; connected sites only through the browser on a machine that is on | A dedicated machine: files, code, terminal, a server-side browser with captured cookies, long jobs in the background, app hosting | Syniaps |
| Knowledge | Projects with files uploaded by hand, Drive and Gmail read on demand, no base built on its own | Synapses: a company wiki built and maintained by the agent, OCR, retrieval, a full cloud mirror, email attachments | Syniaps |
| Setup | An account in 30 seconds, but every task, schedule, Project and Skill is written by you | A short wizard, then the agent proposes the automations itself and the human confirms with a card | Syniaps |
| Learning | Preferences and formats retained as memory files, Skills captured manually | Retains corrections, skills and preferences, and proposes what it has learnt for approval | Syniaps |
| Sovereignty | SOC 2 Type II, ISO 27001 and 27701, but multi-tenant SaaS, data in the United States and Singapore, Chinese roots, and forced data deletion in August 2026 | A dedicated machine in France (OVH), a sandbox per user, secrets outside the database; AI processed in Europe if the company chooses, no certification | Syniaps |
| Entry and price | Free then $20 to $200 a month, self-serve, opaque credits consumed even on failure, Trustpilot 1.2/5 | A 7-day trial open within minutes, a card with no charge before it ends, from €99 excl. tax a month with AI and the machine included | The alternative |
| Scope | A consumer generalist: research, content, slides, websites, apps, design, music, mobile | The company: missions run by eleven expert teams, email, calendar, WhatsApp, LinkedIn, MCP, GitHub, databases | Level |
What each one does better
Where Syniaps wins
A dedicated machine in France
Each customer has their own isolated OVH server, where Manus is a multi-tenant SaaS whose data lives in the United States and Singapore.
Knowledge built on its own
Synapses assemble themselves from documents, email and cloud storage with no manual upload, where Manus waits for you to put files into a Project.
The agent proposes, the human confirms
Automations and missions are born in conversation with card approval, whereas Manus requires you to write every scheduled task and documents no guardrail on sensitive moves.
Sites without an API, computer off
Captured cookies let the agent act on connected sites from the server, where Manus's Browser Operator needs a computer and a browser left running.
Predictable cost
A monthly plan known in advance, with AI included, against Manus credits that are unpredictable, do not roll over and are charged even when the task fails.
Where Manus wins
Immediate, free entry
An account in 30 seconds and 300 credits a day with no card, where Syniaps has no free plan and asks for a card for its 7-day trial.
Massive parallelism
Wide Research launches dozens of sub-agents across hundreds of items, and 20 tasks run at the same time on a Pro plan.
Certifications already obtained
SOC 2 Type II, ISO 27001 and ISO 27701 are published on a trust centre, which Syniaps does not yet have.
Rich deliverables out of the box
PowerPoint slides, a design view, mobile apps, music, ElevenLabs voice and self-published websites all come out of a single prompt.
Ecosystem and name recognition
A Skills library, Playbooks, an API, Zapier, mobile and desktop apps, $125m of annualised revenue and a globally known brand.
Two caveats on its promise
- The promised autonomy meets reality badly: the 2026 reviews report loops, tasks longer than five steps that fall over, "high service load" errors and credits charged on failure, with a Trustpilot score of 1.2/5 dominated by billing complaints.
- Sovereignty is fragile: a company with Chinese roots that passed under Meta then was broken up on Beijing's orders, a compulsory deletion of eight months of data for "certain users" in August 2026, a privacy policy under Singaporean law and storage in the United States and Singapore.
The ten criteria, scored and justified
Intelligence
Frontier models (Claude, GPT, Gemini) and ways to step up a gear
Syniaps 5/5
Claude Sonnet 5, Opus 5 and Fable 5, GPT through Codex, or the model of your choice; Standard, In-depth and Maximum modes.
Manus 3/5
Reasoning rests on the Claude family with Qwen in support, but the customer chooses neither the provider nor the version, only Lite, 1.6 or Max.
Execution power
Actually acts: files, code, browser, sites without an API, long-running work
Syniaps 5/5
Files, code, terminal, a browser for sites without an API, long jobs in the background, hosted apps.
Manus 4/5
A cloud VM with a browser, code and files, parallel sub-agents, deployed websites, Browser Operator and a desktop app, but long tasks fall over (context, loops) and connected sites require a machine that is switched on.
Company knowledge
A base built and maintained on its own from documents, email and cloud storage
Syniaps 4/5
Synapses built by the agent, OCR, a full cloud mirror; still young on very large volumes.
Manus 2/5
Projects with files uploaded by hand and connectors read on demand; no company base built or maintained on its own from documents, email or cloud storage.
Zero setup
Nothing to brief, trigger or tick; the tool proposes, the human confirms
Syniaps 4/5
A short wizard, and the agent proposes its automations in conversation; site connections still have to be made once.
Manus 2/5
No wizard and no spontaneous proposal: every task, schedule, Project, Skill and connector has to be written or wired by the user, and proactive agents remain a roadmap promise.
Continuous learning
Remembers corrections and preferences without being asked to
Syniaps 4/5
Corrections, skills and preferences are retained and proposed for approval; no learning across customers yet.
Manus 3/5
A file-based memory retains formats and preferences between sessions, but Skills are captured manually after a successful task.
Autonomy
Missions and automations server-side, computer switched off, human approval on sensitive moves
Syniaps 4/5
Missions and scheduled actions run server-side with your computer off, with approval cards; mission ceilings still need tuning.
Manus 3/5
Scheduled Tasks run in the cloud with the computer off, with history and alerts, but no human approval is documented before a sensitive move and failures burn credits.
Integrations and connections
Everyday tools, MCP, cloud storage, email, sites without an API
Syniaps 4/5
Email, cloud storage, meetings, WhatsApp, LinkedIn, MCP, sites without an API; fewer native connectors than a Zapier.
Manus 4/5
Native MCP connectors (Gmail, Calendar, Drive, Notion, HubSpot, Stripe, GitHub, Slack), custom MCP, Zapier, an API, Telegram and Browser Operator for sites without an API, but no documented Microsoft 365, WhatsApp or LinkedIn.
Sovereignty and security
EU hosting, per-customer isolation, EU or on-premise inference, certifications
Syniaps 3/5
A dedicated machine in France (OVH), strong per-user isolation and AI processed in Europe if you choose; no on-premise option and no certification.
Manus 2/5
SOC 2 Type II and ISO 27001/27701 are real, but this is a multi-tenant SaaS in the United States and Singapore, under Singaporean law, with inference at Anthropic and Alibaba, Chinese roots and a data deletion imposed by the regulator in August 2026.
Teamwork
Shared spaces, roles, delegation, company-wide memory
Syniaps 4/5
Company sources connected once, shared synapses, role profiles, shared skills and conversations, shared mailboxes in missions; no SSO and no fine-grained roles.
Manus 3/5
Team at $20 a seat with a credit pool, WorkOS SSO, real-time Collab and shared Projects, but task visibility limited to the creator, no fine-grained roles, and session cookies wiped as soon as a guest joins.
Price and access
Total cost (tool + AI + machine) against what you get, and how easily you start
Syniaps 3/5
A dedicated machine, AI included and a 7-day trial that opens within minutes, from €299 excl. tax a month; a card is required for the trial and there is no free tier.
Manus 3/5
Free, self-serve and $20 to enter, but opaque credits that do not roll over and are consumed even on failure, paid SSO below 30 seats and a 1.2/5 on Trustpilot over billing.
Verdict
The deliverable within the hour against the intelligence that carries the company over time
Manus
Manus is the best consumer entry point to an agent that executes, but it starts from scratch on every task, on an unstable SaaS.
Natural fit: freelancers, growth teams, researchers and prototypers who want a report, a deck or an MVP with nothing to install.
Syniaps
Syniaps sells the opposite: a dedicated machine that already knows the company, proposes its missions and acts with the computer off, under human approval.
Natural fit: companies handing over recurring and sensitive missions (email, connected sites, documents) that require isolation and predictability.
Manus sets the market's expectation on free entry and spectacular deliverables, not on knowledge or sovereignty.
Do not chase its demo effects (music, design, mobile) or its free plan: build on what it lacks, knowledge built on its own, a dedicated server in France, execution with the computer off. Make the first hour as smooth as its own (a visible trial, a first result within 10 minutes) and aim early for a certification to turn the security argument around.
Public sources (36)
- https://manus.im
- https://manus.im/pricing
- https://manus.im/team
- https://manus.im/security
- https://trust.manus.im
- https://manus.im/docs/introduction/welcome
- https://manus.im/docs/introduction/plans
- https://manus.im/docs/features/scheduled-tasks
- https://manus.im/docs/features/collab
- https://manus.im/docs/features/projects
- https://manus.im/docs/features/skills
- https://manus.im/docs/features/desktop
- https://manus.im/docs/integrations/integrations
- https://manus.im/docs/integrations/mcp-connectors
- https://manus.im/docs/integrations/manus-browser-operator
- https://manus.im/blog/manus-max-release
- https://manus.im/blog/Context-Engineering-for-AI-Agents-Lessons-from-Building-Manus
- https://manus.im/blog/a-note-to-our-users
- https://manus.im/blog/manus-resumes-independent-operations
- https://help.manus.im/en/articles/16147831-service-change-overview-what-s-happening-and-am-i-affected
- https://help.manus.im/en/articles/12807937-where-can-i-enable-subscribe-to-a-single-sign-on-sso-subscription-for-manus-team
- https://www.marketscreener.com/news/meta-platforms-inc-cancelled-the-acquisition-of-butterfly-effect-pte-ltd-ce7f5cd8da8cf724
- https://www.scmp.com/tech/big-tech/article/3365986/manus-resumes-solo-operations-after-collapse-us2-billion-meta-deal
- https://techstartups.com/2026/09/01/ai-startup-manus-resumes-independent-operations-after-china-kills-metas-2-billion-deal/
- https://www.omm.com/insights/alerts-publications/china-unwinds-meta-s-acquisition-of-manus-implications-for-cross-border-ai-transactions/
- https://theairankings.com/butterfly-effect/
- https://en.wikipedia.org/wiki/Manus_(AI_agent)
- https://www.lindy.ai/blog/manus-ai-pricing
- https://www.nocode.mba/articles/manus-ai-pricing
- https://www.lindy.ai/blog/manus-ai-review
- https://www.eesel.ai/blog/manus-ai-reviews
- https://www.trustpilot.com/review/manus.im
- https://www.g2.com/products/manus-ai-agent/reviews
- https://solvea.cx/blog/is-manus-ai-safe
- https://www.luizasnewsletter.com/p/manus-ai-why-everyone-should-worry
- https://explainx.ai/blog/manus-independent-operations-resumed-september-2026
Manus is a trademark of its publisher, cited here for the purpose of objective comparison. Syniaps is neither affiliated with nor a partner of that publisher. Found an error in this comparison? Write to us at [email protected] and we will correct it.