SignalOS for agencies
SignalOS for agencies is one account for running AI search visibility across several clients. It gives you a measurement allowance shared across your clients, reports and a client portal carrying your agency's name, and an API for your own reporting. Every measured figure carries its sample size and links to the AI answer behind it.
How does the arithmetic work for an agency running several clients?
The agency plan isn't published in our plan catalog right now, so this page shows no figures rather than typing in numbers that could go stale. The shape of the arithmetic doesn't depend on them.
An agency plan is one fixed platform price covering a set number of clients at a time. That price doesn't rise with what you charge your clients, and what you charge them is yours to set, in your own contract. SignalOS never sees it.
Divide the platform price by the number of clients you run on it and you have what each client carries. Every client you add lowers that figure. Put your own fee next to it and the difference is yours to cover your team's time with, which is the real cost of delivery and the subject of the last section of this page.
We don't put a fee into that arithmetic, because we don't know what you charge and a number we invented would be the first thing a careful reader distrusted. Live prices and client limits appear on pricing as soon as they're published.
What does an agency get beyond measuring its own brand?
These are the capabilities that matter when the brands you measure belong to your clients. Each row says what the product does today, not what it might do later, and which published plans include it. Every other capability is listed on features.
| Capability | What it does today | Plans |
|---|---|---|
| Pooled measurement | One monthly allowance of measurement observations — one AI answer to one question on one engine — shared by every client on the account. It's a hard limit, checked when each observation is recorded. As it runs low, a run takes fewer samples per question, and below a floor it takes none rather than report a weaker number. Questions aren't pooled: the only question limit is a ceiling on each query set. | Every published plan |
| Per-client caps | Give each client a monthly allocation of the pool, with optional rollover of what it didn't use, and see every client's consumption ranked against its own cap and the pool. Caps may add up to more than the pool, and the page says when they do. A cap is a planning line, not a hard stop: the pool is the limit that stops measurement. | Specific plans only — none published right now; see pricing |
| Client swap | When a client leaves, one action archives them and creates their replacement. An archived client doesn't occupy a client slot, and its history is kept rather than deleted. If the replacement can't be created, the departing client is restored. | Every plan, within its client limit |
| White-label | Your agency name, logo and colours on the client portal and on report PDFs, and your sender name on scheduled report emails. A custom domain can be recorded but doesn't route to the portal yet, and a few surfaces still name SignalOS — see the first question below. | Specific plans only — none published right now; see pricing |
| Client portal | Five screens a client signs in to: Overview, Trend, This period, What changed and What's next, with a comment thread your team can answer. A client user sees only the clients you grant them. | Specific plans only — none published right now; see pricing |
| API and MCP | A REST API and an MCP server, authenticated with an API key, for your own reporting or AI tooling. Responses carry sample size and confidence where the figure has them, and a link back to the evidence. Access is checked on every request, so a key stops working if the plan no longer includes it. | Specific plans only — none published right now; see pricing |
White-label, the client portal, API access and per-client caps are included on specific plans, not on every plan. No plan that includes them is published in our catalog right now; pricing shows what each plan includes as soon as it is.
What do you say when a client asks how you know?
The risk for an agency isn't a wrong number. It's a client asking how you know, and you not having an answer. SignalOS is built so that every figure you put in front of a client already carries its answer.
A sample size on every figure
AI answers change from one asking to the next, so SignalOS asks each question five times per engine by default and reports a rate with the number of answers behind it. A client who asks "out of how many?" gets a real number.
The AI answer behind the number
Click a figure and the evidence drawer opens: the raw AI answer with the brand mention highlighted, the engine, the model version, the time it ran, and which sampled runs were counted or excluded.
A refusal below the threshold
When there aren't enough answers to support a figure, the product shows no number. It says how many answers it has and how many it needs, rather than a zero or an estimate you'd later have to take back.
Sampling, confidence intervals and the limits of what we can measure are set out in full on methodology, so a client who wants the detail can read it for themselves. Whether AI visibility is urgent for a client's category at all is a separate question, answered on who it's for.
How much of your team's time does each client take?
This is the objection agencies actually raise, so here is the real work, step by step. Setup happens once per client. After that, measurement runs without anyone starting it, and the work arrives as a list sized to the time you say you have.
Setting up a client
1. Domain
Create the client with its name, website and the countries it sells in.
2. Profile
Add the client's products or services and its competitors. A best-effort crawl of the site runs here too, but setup never waits on it.
3. Questions
SignalOS generates a set of buyer questions from that profile. Your team reviews the set and approves it; nothing is measured until you do.
4. First measurement
The first measurement runs against the approved questions, so the client has real data from the start.
The setup wizard keeps no progress record of its own. Each step is worked out from what already exists for the client, so you can close it halfway and come back to exactly the right step.
After setup, measurement repeats automatically on the cadence the plan sets, for every client on an account with an active subscription. Nobody on your team has to start a run.
The capacity-paced worklist is on the plans that include it. You declare, per client, how many hours or items your team has for each period, and the worklist releases only the recommendations that fit. The rest wait in a backlog, and work with an unfinished prerequisite is held back. If no capacity has been declared, nothing is released, rather than two hundred recommendations nobody will action.
Questions agencies ask
- Will my clients know you exist?
- On a plan with white-label, the client portal and report PDFs carry your agency's name, logo and colours, and scheduled report emails use your sender name. It isn't complete yet, and we'd rather you hear that from us: a custom domain doesn't route to the portal, the browser tab icon is still ours, the dashboard snapshot image still says SignalOS, and email subject lines, error and sign-in pages and PDF file properties haven't been audited. If it matters for a particular client, ask in the walkthrough and we'll show you what they would see.
- What happens when a client leaves?
- You archive them, or use the swap action to archive them and create their replacement in one step. An archived client stops occupying a client slot straight away. Its measurement history is kept, not deleted, and your account's full data export includes every client, archived ones too. There isn't a separate export for a single client yet.
- Can I resell at whatever price I want?
- Yes. What you charge your clients is set in your own contracts. SignalOS doesn't see it, doesn't set a minimum, and the platform price doesn't change with it.
- What if one client needs far more measurement than the others?
- The measurement allowance is pooled, so a demanding client can use more of it than the rest. Per-client caps let you plan how the pool is shared and show which clients are consuming it, ranked. A cap is a planning line rather than a hard stop, though: the pool itself is the limit, and once it's used, measurement pauses until the next month.
- How much of my time does each client take?
- Setup is four steps: create the client, add its offerings and competitors, approve the questions SignalOS generates, and run the first measurement. After that, measurement repeats automatically on the plan's cadence, and on plans with the paced worklist the work arrives sized to the hours or items you declare for that client, not as an unbounded list.
- Do I need to be an SEO specialist?
- No. Measurement, citation analysis and diagnosis run from the data, and the work arrives as recommendations. What your team brings is judgement about the client: which of that work fits their business, and how much of it they have capacity for.
Add it to what you already sell.
Book a walkthrough and we'll run it against one of your clients' brands, live, so you're looking at real data rather than a demo.