Cue by Manus gives each AI agent its own email address, phone number, wallet and computer — and that identity is what makes it sellable. A freelancer can now offer a small business an agent that answers its calls, watches for appointments, chases suppliers by email and pays for things inside a budget the owner approves. This guide turns those capabilities into five service packages with clear deliverables, plus the guardrails you need before a client’s name is on anything.
The honest caveat comes first: Cue is still in invite-only early access, launched 28 September 2026, and we have not hands-on tested a live account — everything here rests on Manus’s own materials and independent coverage read on 22 October 2026. Treat the packages as sellable now, but price and deliver around the product’s rough edges, not its launch slide deck. If you have not set up Cue yet, read our Cue by Manus setup guide first; the Manus Cue explainer covers what the app is.
Why a Cue agent is worth paying for
A client is not buying “AI” — they are buying answers they never had to chase.
That distinction decides whether anyone pays you. A normal chatbot answers questions. A Cue agent has the identity pieces to act: an email that can write to a supplier, a phone number that can take a call while the owner is in a meeting, a wallet that can hold a deposit inside a budget, and a computer where multi-step work finishes on its own. Manus’s own examples make the pitch concrete: the agent takes a forwarded call and files a summary, checks every day for an earlier driving-test or dermatologist slot and books it the moment one opens, or writes to several hotels asking which can guarantee connecting rooms and brings back their answers.
Two pricing facts shape the business model. Cue itself is free during early access with an invite code, and Manus has published no price for after it — so your cost base today is effectively zero, but a client who depends on the service must accept that the free period ends eventually. The phone number is the exception: it is a paid subscription bought separately from any Manus membership, with phone use counting toward a weekly Cue allowance of unpublished size. A call-answering service needs that number, so price the service to cover a cost you will have to discover in the app. (Source: CellCog, Cue by Manus: What It Is, Invite Code and Price; HokAI, Is Cue by Manus Free?; both read 22 October 2026.)
The 5 sellable service packages
Each package below is one job, one measurable deliverable, and one permission level — start every client at the lowest rung that works.
Our AI agent permissions guide sets the ladder: Rung 0 (read-only research), Rung 1 (drafts for your approval), Rung 2 (capped, with your confirmation on every spend, call or sent message), Rung 3 (bounded autonomy). For client work, Rung 2 is the ceiling until the agent has proved itself for weeks. Keep irreversibles at approval-on forever.
1. Missed-call answering and triage
The deliverable: every missed call answered, summarised, and filed by urgency — a “land, then the rest” briefing.
Ideal client: tradespeople, clinics, small agencies — anyone who loses jobs to the voicemail they never return. Manus’s own example is the script: the agent answers, finds out who called and why, and when the owner checks in, it reports the airline and the hotel first, then the rest sorted by urgency. That triage is the product; the call answering is just the mechanism.
What it needs: a Cue agent with a phone number — the paid add-on — so price this package to absorb a subscription of unpublished size, and confirm the country’s phone support first (Manus says phone availability varies by country; some locations support voice calls only). Start at Rung 1: the agent drafts summaries and never calls out without approval. Get the client’s written consent to an AI answering their business number before anything is configured.
2. Appointment and slot monitoring
The deliverable: a booked slot the client could not get themselves — or a daily “no opening yet” report.
Ideal client: driving schools, clinics, campsites, ticketed venues — anyone whose customers fight over scarce slots. Manus’s own use cases cover this exactly: checking daily for an earlier driving-test slot, a dermatologist appointment or a full campsite, and booking it the moment one opens. Cue’s automations can also trigger from events, not only schedules — for example, a fare dropping below a set level.
What it needs: Rung 2 with payment confirmation on. Before paying, Manus says the agent shows the times, prices and change/cancellation policies and waits for confirmation, paying only inside the budget the owner sets. The client, not you, confirms each booking; your job is configuring the watch and reading the daily report.
3. Outreach and follow-up by email
The deliverable: a shortlist with real answers — suppliers, hotels or partners who replied, ranked with their terms.
Ideal client: event planners, travel agents, procurement-light businesses. This is Manus’s hotel example made commercial: the agent writes to several hotels asking which can guarantee three connecting rooms and returns their answers. You sell the legwork clients hate — first contact, chasing replies, compiling the comparison.
What it needs: the agent’s own email address, used transparently. Disclose to every recipient that they are corresponding with an AI agent on the client’s behalf; a reply chain built on a misled supplier is a liability, not a service. Run at Rung 1 until the client approves the outreach template, then Rung 2 for the chasing, with sends logged and reviewable. Never run outreach from the agent’s identity on accounts it was not given.
4. Event and venue research shortlists
The deliverable: a venue shortlist plus a draft deck — research done to presentation-ready.
Ideal client: anyone planning a launch, wedding or corporate event. This is the Cue group-chat demo: several agents in one chat with a shared goal — one researches venues for a New York-style launch, another builds the shortlist, a third drafts the deck — while the client sets the direction and makes the final call. You sell the output, not the chat.
What it needs: Rung 0 to start (pure research, nothing sent), moving to Rung 1 for the deck draft. No money moves, no client accounts connected. Because group-chat handoffs can drift — one independent test found an agent that acknowledged a task but needed a second explicit instruction before producing it — read every handoff in the log before the deliverable leaves your hands. (Source: WandaBuilds, Manus Cue: AI agents that take your calls, read 22 October 2026.)
5. Travel-itinerary planning package
The deliverable: a day-by-day itinerary, packing list and organised bookings in one handoff.
Ideal client: busy professionals, families, honeymoon planners. The Cue group-chat example is the service: one agent builds the three-day Kyoto-style itinerary, another makes the packing list, a third organises the tickets — handing work to each other while the client reviews. Package it as a fixed-fee deliverable with one revision round.
What it needs: Rung 1 for the plan, Rung 2 (approval on, budget capped) for any booking. The traveller confirms every payment. Keep Memory and Dream features — Cue’s persistent memory of preferences — scoped to that client only; do not let one client’s tastes bleed into another’s agent.
| Package | Deliverable | Starting rung | Paid extra needed? |
|---|---|---|---|
| Missed-call triage | Urgent-sorted call summaries | Rung 1 | Yes — agent phone number |
| Slot monitoring | Booked slot or daily “no opening” report | Rung 2, approval on | No (booking budgets are the client’s) |
| Outreach & follow-up | Ranked replies with terms | Rung 1 → Rung 2 | No |
| Venue research shortlist | Shortlist + draft deck | Rung 0 → Rung 1 | No |
| Travel itinerary | Itinerary + packing list + bookings | Rung 1 → Rung 2 | No |
Guardrails before you take a client’s money
Get consent in writing, keep payments approval-on, and keep client data inside the client’s accounts — these are the non-negotiables.
- Consent for the phone and wallet. The agent calls and pays in the client’s name. Written consent covers: that an AI answers their calls, that it may hold a funded wallet with a stated cap, and that payments need their confirmation. Manus has not documented the wallet’s payment rails, refund process or transaction audit trail — say that to the client before you sell anything with money in it.
- Keep identities separate. Each Cue agent gets its own email, number, wallet and computer “depending on availability”, per Manus’s own listing — but a separate identity is not a data boundary. Use one agent per client, review connected apps per agent, and never reuse an agent across two clients. Connected-account and context checks are the boundary, not the agent’s name.
- Disclose the AI. Anyone the agent contacts should know they are dealing with an AI agent on the client’s behalf. It costs nothing and converts a trust risk into a talking point.
- Log everything. Read the agent’s activity trail daily during the first weeks. The honest test from our setup guide still applies: does the agent contact the right person, spend on the right thing, and stop when it should? If the trail is unreadable, do not widen permissions.
- Watch for prompt injection. Agents that read email can be targeted by hidden instructions in messages they read — Salt Labs disclosed a since-patched flaw of exactly this kind on 24 September 2026. Keep permissions narrow and approvals on.
What to charge: honest pricing guidance
Charge for the deliverable, not the tool — and do not invent income figures that do not exist yet.
There are no published earnings reports from Cue freelancers; the product is weeks old and invite-only. What you can price honestly is the job: a fixed fee per deliverable (a completed shortlist, a month of call triage, a booked itinerary) beats hourly billing because the client buys an outcome, and you bear the automation risk. Factor in the real costs: the phone number is a paid Cue subscription of unpublished size for call-heavy services, and the weekly usage allowance is also unpublished — find both in the app before quoting a retainer. Be explicit with the client that Cue is in early access and terms may change; writing a “terms may shift” clause into your agreement is cheaper than an argument later.
Honest limits before you sell
Three facts bound everything above — sell inside them.
First, access is still invite-gated. Manus’s launch code MEETCUE went to a limited pool on a first-come basis, existing Manus users enter through the Agents tab, and everyone admitted receives extra codes to share — but nobody can just download and log in. You can sell services only as far as your invite codes stretch. Second, the full identity set is not guaranteed: “depending on availability, an agent can have its own phone number, email address, wallet, and dedicated computer” is Manus’s own phrasing, so verify each agent actually holds what the package needs. Third, the product is young: Google Play reviews average 4.7 but one-star reports describe beta-state crashes, slowness and invite-code errors, and an update landed as recently as 19 October 2026 (v26.18.1). Sell the service as early access, staffed by a human — you.
Frequently asked questions
Can you really sell services built on Cue by Manus?
Yes, as a freelancer selling deliverables, not as a reseller of the app. The model is you operating the agent and selling call triage, slot monitoring, outreach, research shortlists or itinerary planning to small businesses. Get client consent for the phone and wallet, keep payments approval-on, and price fixed-fee per outcome.
Is Cue by Manus free for freelancers?
During early access, yes — with an invite code. Manus has published no post-early-access price. The phone number is a paid add-on bought separately from any Manus membership, and phone use draws on an unpublished weekly allowance, so call-heavy services carry a real cost you must discover in the app.
How do you get a Cue invite code?
Manus’s launch code MEETCUE (first-come, first-served, possibly exhausted), the Agents tab inside the Manus app (needs a Manus account), or an extra code from someone already inside. Full details are in our Cue setup guide.
What is the safest first client package?
The venue research shortlist. It runs at Rung 0–1: pure research and a draft deck, no client accounts connected, no money moving, nothing sent without approval. Prove reliability there before selling call answering or payment-adjacent work.
What are the risks of client work on an early-access agent?
Invite-only access, identity pieces that are not guaranteed, an undocumented refund process, beta-state bugs, and prompt-injection exposure for email-reading agents. Disclose all of them to the client in writing and keep a human — you — confirming anything irreversible.
Sources and methodology
- CellCog — Cue by Manus: What It Is, Invite Code and Price (invite routes, phone subscription separate, “depending on availability” identity caveat, free early access; read 22 Oct 2026)
- HokAI — Is Cue by Manus Free? Invite Access, Limits & Pricing (2026) (unpublished phone price and weekly allowance, spend comes on top; read 22 Oct 2026)
- WandaBuilds — Manus Cue: AI agents that take your calls, send emails and pay for you (Manus’s own examples: call triage, slot booking, hotel outreach; approval-before-payment framing; read 22 Oct 2026)
- AppBrain — Cue by Manus (v26.18.1 updated 19 Oct 2026; 4.71 rating, mixed reviews; read 22 Oct 2026)
- OpenAIMaster.ai — Manus Cue Explained (launch facts, Salt Labs 24 Sep disclosure; own reporting)
- OpenAIMaster.ai — Cue by Manus Setup Guide (permission ladder, invite routes, free-limits checklist; own reporting)
Methodology: this article rests on Manus’s 28 September 2026 launch materials as reported by the independent sources above, plus Cue’s Google Play/App Store listings and release history, all re-checked 22 October 2026. Early access is invite-only, so no hands-on testing was possible; the five service packages are our original desk-built framework, every capability is framed as what Manus says, and every unverified figure is labelled. Internal links point to our own reporting for follow-up reading.
Arva Rangwala covers AI news, AI tools, guides and prompts for OpenAIMaster — what is new, what is worth using, and how to put AI to work.