Skip to content

Agencies

Bring your own data. Sell it as yours.

You already own the client relationship, the data and the trust. Parleon is the layer underneath it: the protocols, the endpoint, the pages and the measurement, running under your brand at wholesale pricing.

Bring your own

Whatever your data lives in

  • A feed on a URL or SFTP mapped
  • An API nobody else supports mapped
  • A database you own read only
  • A spreadsheet someone maintains on a schedule

Describe the source once, map the fields once, and it becomes tools an assistant can call. Under your brand.

The question every agency asks first

What happens to my data?

Every agency arrives with something specific: a feed they negotiated, a client's booking system nobody else integrates with, a warehouse, a spreadsheet one person maintains. A platform that can only ingest what it already knows about is a platform you cannot sell, because your best clients are exactly the ones with the strangest systems.

So the ingestion side is deliberately dumb and general. You describe the source once, map the fields once, and it becomes tools an assistant can call. Nothing about the shape of your data has to match ours.

Anything that can answer a question can back a tool.

A feed you already have

CSV, XML, JSON, on a URL or an SFTP drop. Point us at it, tell us what the columns mean once, and it becomes tools an agent can call.

An API nobody else supports

Your client's booking system, their PMS, their warehouse, the thing their cousin built in 2015. If it answers over HTTP, it can sit behind an MCP tool.

A database you own

Read-only credentials against Postgres, MySQL or a warehouse, with the query surface you define and nothing wider.

A spreadsheet somebody maintains

Not glamorous and extremely common. A Sheet or an Excel file on a schedule is a perfectly good source, and it beats the answer a model invents.

The client's own website

Where there is no system at all, we read the site, structure what is there, and tell you honestly which fields we could not find.

Whatever comes next

New sources are an adapter against the same brain, not a rebuild. If you bring one we have not seen, the work is a connector, not a project.

Custom tools, not just custom data. A source becomes tools, and you decide what those tools are called, what they take and what they return. If your vertical needs a tool that exists nowhere else, you define it once and every client on your book has it.

The console

Onboard, launch, update, keep.

Four things, and the fourth is the one that decides whether you keep the account.

01

Onboard

Add a client, point at their data, and the layer builds itself: discovery documents, an MCP endpoint, WebMCP tools on their own domain. No ticket, no engineer, no waiting on us.

02

Launch

One path on their hostname points at the edge and the layer is live. Their website vendor is not involved and their CMS is not touched, which is what makes this sellable to clients who do not control their own site.

03

Update

Change the source, change the tools, change the copy, push to the whole book at once or one client at a time. Versioned, so a bad change is a rollback rather than an incident.

04

Keep

Freshness monitoring, per-client visibility tracking, and an alert when a feed goes stale or an assistant starts getting a client wrong. Retention is a monitoring problem, not a relationship problem.

Bulk actions are the point. Pushing a change to sixty clients has to be one action, or the eightieth client is the one that quietly goes stale.

White label

Your name on all of it.

The client should not have to know we exist. Nothing we serve on their domain carries our name, the reports come from you, and the console can wear your brand.

  • Your logo, your colours and your domain on the console and the reports
  • Client-facing pages served from your client's domain, never ours
  • Reports that go out under your name on your schedule
  • Your pricing to your clients, wholesale from us, no floor we enforce
  • Your support in front, ours behind, and a route to an engineer when it is genuinely ours

How the economics work

Wholesale

You buy at a rate that drops as the book grows, and you set retail. We do not publish a floor and we do not undercut you to your own client.

One invoice

One bill for the whole book, per client line items, so your reconciliation is a spreadsheet rather than a project.

No minimum to start

Put one client on it. If it does not hold up, you have lost a month rather than a commitment.

The awkward questions

Whose customer is it?

Yours. We do not sell to your clients, we do not contact them, and if you leave, you leave with them. The wholesale agreement says so in the boring part.

What if a client's data is a mess?

That is the normal case. We tell you which fields are missing rather than filling them in with something plausible, because a confident wrong answer is what this whole product exists to prevent.

Can we build our own tools on top?

Yes. Every client gets an MCP endpoint you can call, and you can define your own tools against your own data. If you want a tool nobody else has, that is the point of the platform rather than an exception to it.

How long does a client take to launch?

A clean source and a cooperative DNS change is same-day. The long pole is almost never us; it is getting someone to point one path at the edge.

Put one client on it and see.

Bring the messiest data you have. That is the useful test, and it is the one we would rather fail early.

Free, about a minute, no account. Or book a demo.

Enter your domain and see what an AI can read.

Free, about a minute, no account.