Manus AI
GAMMA-Ray report · Spectrum 0.2 beta · Public tier · read 2026-11-10 · not reviewed by the company yet
Manus AI makes a general AI agent that plans and carries out tasks for a user in a cloud sandbox, a desktop app, a browser extension and through an API.
of 13 found
and hashed
in 8 of 8 instruments
at work, with sources
Built only from Manus AI's own public pages (documentation, API documentation, security page, blog, customer stories) and, for company history only, Wikipedia. Manus AI has not seen this report. Nothing was tested and no credentials were used. If you are Manus AI and something here is wrong or missing, the correction will be made and dated, and you can claim the card to add evidence.
The agentic solutions
One card per solution. Each level is computed from the evidence by a published rule; a dimension with nothing public reads "not yet evidenced", which is not a no. A card rates a product as documented, not anyone's deployment of it.
Manus (general agent; modes Manus 2.0 Max, Manus 2.0, Manus 2.0 Lite)
Agent that plans a task and runs each step in its own cloud sandbox with internet access and a file system, delivering reports, slides, websites and other outputs.
Standout. Autonomy 3 (Acts): "tools that the current Manus agent runs in the main task loop"
Next unlock. Control 2 (Configurable): public evidence that what it can reach is enumerated: scopes, permissions, tools would get there.
Trust. FACT-Ray: 18 facts: 18 π Moderate · found by an agent, captured and hashed, checked by 2 other agents (same model family, shared repository); no human has signed it.
Standards constellation. 52 provisions lit across 8 of 8 instruments.
Evidence: 12 sources, captured and hashed · the card's page has the vault's read key · Not reviewed by its vendor yet · correct or claim this card
The full card, every answer and its evidence · the governance layer
Manus Flex
Runs the Manus agent with the user's own API key from a listed inference provider. No card yet (beta): this solution has not been assessed.
Wide Research
Main agent splits a task into sub-tasks, each handled by a separate agent with its own context, and assembles the results. No card yet (available): this solution has not been assessed.
Cloud Browser
A cloud browser the agent operates; the user may log into accounts in it so the agent acts in them. No card yet (available): this solution has not been assessed.
Browser Operator (My Browser)
Browser extension that lets the agent act in the user's local browser with existing sessions, per authorised session. No card yet (available (Chrome, Edge recommended)): this solution has not been assessed.
Manus Studio (desktop app: My Computer, Computer Use, Remote Control)
Desktop app for macOS and Windows that lets the agent work with local files, terminal and apps under a chosen permission mode. No card yet (available): this solution has not been assessed.
Connectors, MCP connectors and custom MCP servers
Account connections and MCP servers that let the agent read data from and take actions in third-party apps. No card yet (available): this solution has not been assessed.
Manus API v2 and Open App (OAuth2)
Programmatic task creation, polling, action confirmation and webhooks; Open Apps act for team users through scoped OAuth2 tokens. No card yet (available (Open App requires a Team account)): this solution has not been assessed.
Automations and schedules
Runs tasks on a schedule or when something happens in a connected app. No card yet (available): this solution has not been assessed.
Manus for Slack
Users tag @manus in a Slack thread and the agent works on the thread's request. No card yet (available): this solution has not been assessed.
Mail Manus
Tasks are started by forwarding email to a personal Manus bot address, with results returned by email. No card yet (available): this solution has not been assessed.
Manus for Zapier
Triggers Manus tasks from events in other apps through Zapier workflows. No card yet (available): this solution has not been assessed.
Website and app builder (Manus 2.0 web development)
The agent builds, hosts and publishes websites, web apps and mobile apps. No card yet (available): this solution has not been assessed.
Our reading
The shape. Manus reads A3 · R2 · C1 · O2 (Building). The API documents the actions the agent carries out itself: sending email, changing calendar events, running commands, deploying a service. With connectors it acts in apps such as Gmail and Notion, and in Manus Studio it handles routine local actions and asks the person when a decision is needed.
Where the evidence is strongest. Autonomy is backed by a named customer: k-ID's team runs more than eight thousand Manus sessions, by Manus's own account. The API publishes a list of action types that wait for a person's confirmation, and an endpoint to stop a running task.
Control at level 1, and what would lift it. The confirmations are documented for the API, which is level 1, marked available because the page does not say when they apply in the app. Level 2 asks for access narrower than the whole account; Manus Studio's approval modes limit what the agent changes, not what it reads. Autonomy 4 would need a public statement that the person approves only critical decisions.
Openness. Limits (for the Browser Operator) and data use are documented, which meets Disclosed and Documented under Spectrum 0.2. Manus lists SOC 2 and ISO 27001 on its security page; the certificates sit behind a page we could not read without JavaScript, so they are shown and not counted.
For readers. The pages tell their own history: Meta acquired Manus on 29 December 2025, and Manus states it formally resumed independent operations in a post dated 1 September. The customer stories are Manus's own and are reported here as such.
What would move it
The one piece of public evidence that would lift each card next:
- Manus (general agent; modes Manus 2.0 Max, Manus 2.0, Manus 2.0 Lite): Control 2 (Configurable). Public evidence that what it can reach is enumerated: scopes, permissions, tools.
The agent at work
Examples from public pages that name a customer or give figures. Each source was captured and hashed on the date shown.
k-ID
Company customer story: k-ID's team uses Manus for legal and policy research and client reports; states session count and spend. Vendor-published. "k-ID now runs over eight thousand Manus sessions across the team."
Source (captured 2026-11-10T04:59Z, sha256 0f7caa90d1e9b97a)
Ascendea
Company customer story: Ascendea uses a Manus agent it calls 'James' at the top of a three-layer agent system for complex work, with human review. Vendor-published. "Ascendea placed Manus at the top of a three-layer system"
Source (captured 2026-11-10T04:59Z, sha256 8df8996ccf52f3a7)
Circa
Company customer story: Circa's founders built and hosted the company website and internal tools on Manus and prepared applications with it. Vendor-published. "They built and hosted the company’s website on Manus"
Source (captured 2026-11-10T05:00Z, sha256 3cb07e6d59ef6ac6)
Heicoders Academy
Company customer story: Heicoders built operational systems such as a hiring portal with Manus. Vendor-published. "Within months of adopting Manus, Heicoders rebuilt their entire operational infrastructure."
Source (captured 2026-11-10T05:00Z, sha256 c1cbca96f97b53d2)
Mandai Smoke Co.
Company customer story: the owner built an order prep calculator and logistics workflow with Manus. Vendor-published. "He asked Manus to build a simple prep calculator"
Source (captured 2026-11-10T05:00Z, sha256 8a407b9e00e695d7)
Manus (company figures)
Company-stated usage figures for the agent: tokens processed and virtual computers created since launch. "Manus has processed more than 147T tokens and created more than 80M virtual computers."
Source (captured 2026-11-10T05:00Z, sha256 3b1f416135395e5c)
The company
- The company states it operates from Singapore. "The company will continue to operate from Singapore." Source (captured 2026-11-10T05:00Z, sha256
a5886ffd94de30b0) - The company states Meta acquired Manus on 29 December 2025. "On December 29, 2025, Meta acquired Manus." Source (captured 2026-11-10T04:59Z, sha256
c275abc0a58b774c) - The company states it formally resumed independent operations (post dated 1 September). "Manus has formally resumed independent operations." Source (captured 2026-11-10T05:00Z, sha256
a05d2e825e0c8e39) - The company states it raised $75M led by Benchmark before launch. "Prior to our launch, we raised $75M led by Benchmark ." Source (captured 2026-11-10T05:00Z, sha256
3b1f416135395e5c) - The company stated a headcount of 105 across Singapore, Tokyo and San Francisco (post dated 17 December). "Today, we're 105 people across Singapore, Tokyo, and San Francisco." Source (captured 2026-11-10T05:00Z, sha256
3b1f416135395e5c) - The company states user data is stored in the US and Singapore. "Our data is stored in the US and Singapore." Source (captured 2026-11-10T04:59Z, sha256
c275abc0a58b774c) - Wikipedia states Butterfly Effect was founded in China and is based in Singapore. "developed by Butterfly Effect, a company founded in China and based in Singapore" Source) (captured 2026-11-10T04:59Z, sha256
4de2f3ed3e1ec47a) - The security page lists SOC 2 Type 2, SOC 2 Type 1, ISO 27001:2022 and ISO 27701:2019. "SOC 2 Type 2 SOC 2 Type 1 ISO 27001:2022 ISO 27701:2019" Source (captured 2026-11-10T05:00Z, sha256
096edaae3a819e40)
The people
Public and professional only, from the company's own pages or a public professional source.
- Xiao Hong, CEO. Source
- Tao Zhang, Co-founder and Chief Product Officer; Web Summit Lisbon 2026 speaker. Source
- Yichao 'Peak' Ji, Co-founder and chief scientist (per Wikipedia); author of the context engineering post. Source
- Chetan Puttagunta, Board member (Benchmark General Partner). Source
The graph
The company, its people and the stories on the left; each solution, the captured sources behind its card, the indicators those sources evidence and the provisions of the instruments those indicators touch, to the right. "Touched" never means "complies".
The evidence
How this report was made
A researcher-assessor gathered and captured the sources and mapped the evidence to indicators, without assigning a level. A challenger argued against every level three and four and every doubtful credit. The deterministic scorer computed the levels and the gate checked the rules. A card appears here only after two rounds of review and a sign-off by agents that did not write it (the team's workflow); Manus (general agent; modes Manus 2.0 Max, Manus 2.0, Manus 2.0 Lite): reviewed in 3 rounds, signed off by gamma.approve@web-summit.dinis on 2026-11-10. A person's sign-off of the series is still to come.
| Role | Who |
|---|---|
| researcher-assessor | a sub-agent of gamma.studio@web-summit.dinis |
| challenger | a sub-agent |
| scorer | build/spectrum.mjs |
| analyst, designer, writer | gamma.studio@web-summit.dinis |
| review | gamma.review@web-summit.dinis, 3 rounds |
| sign-off | gamma.approve@web-summit.dinis, 2026-11-10 |
The vault
Every capture, the dossier, both assessors' mappings, the challenger's notes and this report's data are in a public sgit vault (wcvaw1bi). Read key:
sgit_public_read_5ab82ed8d04719fb9fa7ffa28ef3cbd9eec71b4b03b2958ee62f470c3c1e6ba9:wcvaw1bi
With sgit: sgit clone followed by the key. The key reads; it cannot write.
Corrections
None yet. A correction is made in place, dated here, and carried to the card, the vault and the article.
The article
Manus on the GAMMA-Ray Spectrum: A3 · R2 · C1 · O2, written from this report for PrompTugal.news.