Connect Supply Finder to Claude: the MCP Connector

Last updated: September 17, 2026

Supply Finder now has an MCP connector for Claude, so a buying team can ask about pacing, deal fill and site lists in plain language instead of opening a dashboard. Here's what it does, what it doesn't, and how to connect it.

Most people running a Supply Finder campaign already know the questions they ask every day: is this campaign on pace, why isn't this deal filling, which domains are dragging viewability down. Getting the answer usually meant opening the app, finding the right campaign, and reading a dashboard built to show everything at once.

The MCP connector puts those same questions inside a Claude conversation. Ask "is my Q3 Brightwell Foods campaign on pace" and Claude finds the campaign, checks pacing, and answers in a sentence , with the numbers behind it if you want them.

This is written for two audiences at once: the buying team who'll actually use it, and the IT admin deciding whether to approve it. Both questions get answered here , what it does, and what it's allowed to see.

What MCP is, in one paragraph

MCP (Model Context Protocol) is an open standard that lets an AI assistant call tools hosted by another company's server, inside the conversation, instead of the person switching to that company's app. Claude's connector directory is full of these , this is Supply Finder's. Connecting one doesn't hand the assistant a copy of your data; each tool call is a live, scoped request answered by Supply Finder's own backend, using your own account's permissions.

What you can do with it

Ask about a campaign or deal without knowing its id. "How's my SSP deal doing" or "give me the status of my campaign" both work , Claude looks it up by name first (fuzzy, typo-tolerant matching) and asks which one you mean if more than one matches. You never need to go find a campaign_id yourself.
Check delivery and pacing. Ask whether a campaign is under-pacing, on pace, or over-pacing, and get the required run rate to catch up, broken down by deal.
Diagnose why a deal isn't filling. Instead of staring at bid_requests and win_rate yourself, ask why and get the single most likely cause , no bid requests being sent, a low win rate, a low bid rate , in plain language.
Find underperforming inventory. Ask which domains or apps in a deal are dragging viewability, CTR or completion rate below target, and get pruning candidates ranked by how much volume they'd cost you to remove.
See what the optimization engine already proposed, and why. The daily optimization job's own reasoning , not a re-derived guess , for why it did or didn't propose a change.
Query raw delivery metrics. Impressions, clicks, viewability rate, win rate and more, broken down by date, domain, device, deal or SSP, over any date range.
Build and read site lists. Generate a ranked list of domains, apps or CTV bundles from a plain-language brief ("US, 500 sites, no MFA"), then read it back, look up catalog detail on specific domains, or delete a list you no longer need.
Check your followed-campaigns board. The same traffic-light view (spend, impressions, viewability, completion, CTR, margin, each against target) as the "Followed campaigns" panel in the app , without opening it.

What it can't do

The reporting side is read-only by construction: none of those tools can approve a proposal, adjust a bid, or edit a campaign or deal , there's no such tool exposed. The two tools that do change something are narrow and deliberate: generating a site list creates a new saved list (and spends a credit), and deleting one requires a separate confirmation step so nothing gets removed on a single ambiguous request.

What an IT admin should know

A few things are true of every tool in this connector, not just some of them:

Identity comes from the login, not from a parameter. No tool accepts an organization id or account id as an argument. The server resolves who's calling from the authenticated session itself, so there's no field a model could fill in , correctly or by mistake , to reach past the connected account.
Access mirrors the app exactly. A regular org member sees their own campaigns and lists through the connector, the same as in the app. An org owner or admin additionally sees their organization's shared campaigns, again matching what they'd see logged into Supply Finder directly. There's no broader or connector-specific scope.
Metrics queries are closed, not open SQL. The model chooses what to ask for from a fixed list of metrics and dimensions, each with one hardcoded formula; nothing resembling a raw query is ever built from user or model input.
Ids are always re-validated server-side. Passing a deal or campaign id that doesn't belong to the caller's account fails before any data is touched, rather than trusting whatever the model supplies.
One tool spends a credit; everything else is free. Generating a site list draws down the organization's credit balance like it would from the app. Reading data , pacing, fill diagnostics, metrics, existing lists , never does.

How to connect it

In Claude, open Customize in the left sidebar (or, from a chat, click the + beside the message box and choose Connectors).
Go to Connectors and choose to add a custom connector.
Enter the connector URL: https://api.supply-finder.com/mcp/
Click Connect, then log in with your existing Supply Finder account when redirected. There's nothing to sign up for separately , it's the same login you already use in the app.

Once connected, just ask. You don't need to name the tool or the connector explicitly , mention a campaign, a deal, or ask for a site list, and Claude will reach for the right one.

Example prompts to try

"Is my Q3 Brightwell Foods campaign on pace?"
"Why isn't the SSP deal on my Northwind Outdoors campaign filling?"
"Which domains in that deal are hurting viewability the most?"
"Build me a 500-site US web list, exclude anything flagged as low quality."
"What's my followed campaigns board looking like this week?"
"Compare CTR by device for this campaign over the last 30 days."

A first answer is usually the start of the conversation rather than the end of it , once you have a pacing number or a list of underperforming domains, it's natural to keep going: ask why, ask what to do about it, or ask for the same check on a different campaign.

Frequently asked questions

What is the Supply Finder MCP connector?

It's a connector that lets Claude call a set of Supply Finder tools directly inside a conversation: building and inspecting site lists, and reading campaign pacing, deal fill, segment performance and delivery metrics. MCP (Model Context Protocol) is the open standard Claude uses to talk to external tools like this one, the same mechanism behind connectors from other companies in Claude's directory.

Does the connector expose write access to campaigns or deals?

No. All nine campaign reporting tools are strictly read-only , none of them can approve an optimization proposal, change a bid, or edit a campaign or deal. The only tools that change anything are generate_site_list (creates a new saved list) and delete_site_list (which requires a separate confirmation step before it deletes).

What data can Claude see through this connector?

Only what the logged-in user could already see in the Supply Finder app: their own campaigns, deals and site lists, plus their organization's if they're an owner or admin. Every tool resolves the caller's identity from the authenticated session , there's no parameter a model could set to pull another account's or another organization's data.

Does using the connector cost anything?

Generating a new site list consumes one credit from your organization's balance, the same as generating one from the app. Every other tool , reading an existing list, checking pacing, running a metrics query, listing campaigns , is free to call.

How do I connect it?

In Claude, open Customize in the sidebar, go to Connectors, add a custom connector pointing at https://api.supply-finder.com/mcp/, and log in with your existing Supply Finder account when prompted. No separate API key is needed for this flow.

Can an IT admin restrict what the connector can access?

Access follows the permissions the connecting user's Supply Finder account already has , an org owner or admin sees org-shared campaigns through the connector, a regular member sees only their own. There is no elevated or connector-specific scope beyond what that account can already do in the app.