Audited 2026-11-08 · Sources: Manus 2.0 announcement + cue.im product pages, App Store/Google Play listings (Butterfly Effect Pte. Ltd.), personalagentbench.com landscape record, aiinvitecode.com invite tracking, hands-on community reports (note.com, stayingahead newsletter). Tabletop audit — no account created, no agent deployed.
The card
| Vendor | Butterfly Effect Pte. Ltd. — the company behind Manus; Cue shipped as part of the Manus 2.0 release on 2026-09-28, with the iPhone app live on the App Store 2026-10-08, plus Google Play and Microsoft Store builds. |
| What it is | A standalone personal-agents app where each agent gets its own identity — its own email address, phone number, wallet, and cloud computer. It takes the calls you forward to it, emails from its own address, pays within a budget you set, and keeps working after you log off. |
| Access model | Early access, free with an invite code; Cue Plus included during early access (listed in-app value ~$20/mo, pricing checked 2026-11-08 via App Store listing and Manus blog). Launch code MEETCUE was capped. |
| Execution surface | Cloud-only. Agents run on Manus's cloud computers — there is no local install path. Your agent's mail, calls, files, and wallet activity all live on their infrastructure. |
| Control model | Per-agent permissions and approval boundaries: you set the responsibility, connect the services, define routines, and the agent brings work back when a decision needs you. Multi-agent group chat included. |
Why this one matters
Cue is the first mass-market product to ship the architecture serious operator fleets already run internally: agent identity as a first-class object. Separate email, phone, wallet, and machine per agent is how you scope blast radius — one compromised agent can't impersonate the others, and every action is attributable to a distinct identity. That is the model, packaged for consumers.
We have run it against real work. In our own operations this week, a Cue agent was the recon layer that surfaced internal material exposed on our public site — it found the exposure before our own review stack did. As a discovery and verification agent, it is extremely strong.
Our gate findings
| Gate | Finding |
|---|---|
| Upstream identity | Verified — Butterfly Effect Pte. Ltd. on the store listings, same infrastructure as Manus per the vendor's own release notes. Real company, real product lineage. |
| Claims discipline | Honest. "Early access," invite-gated, free-for-now pricing clearly marked as subject to change. No guaranteed-outcome language in the product copy. |
| Capability | Top of consumer class. Persistent agents with identity, connectors, routines, and approval gates — it does real multi-step work, not chat. Owner verdict in production: extremely strong. |
| Data residency | The wall. Everything executes on Manus cloud computers — agent memory, connected accounts, call audio, wallet transactions. No local deployment, no self-hosted option, no data residency control. Your identity graph lives on their servers. |
| Exit cost | High. The agent identities, routines, and memory are platform artifacts. You cannot export an agent — you rebuild it elsewhere. Classic vendor gravity well, softened only by the current $0 price. |
| Blast radius | Bounded by design, total by architecture. Per-agent wallets/phones limit what a single agent can do — good. But the account-level blast radius is everything you connect: a platform breach or policy change hits all your agents at once. |
| Security scan | N/A — closed-source SaaS, nothing local to scan. Evaluation is contractual and architectural, not technical. |
Verdict: GO — as a capability agent and recon layer; NO-GO as your system of record. Cue is arguably the strongest consumer agent shipping today for research, verification, and multi-step work. But it is rented sovereignty: the identities, the memory, and the execution all belong to the platform. Use it for what it does best — discovery, drafting, verification — and keep your own operations on infrastructure you own. If your agent can't run on your metal, you don't own the agent.
Best for
Anyone who wants to see what a real personal agent looks like in 2026 — and every operator who needs a second set of eyes on their public surface. Just remember: the agent that finds your exposure for you can find it for anyone.