MCP Use Cases

Once your assistant is connected, you talk to it in plain language and it picks the tools. You don’t need tool names. What helps is being specific about which campaign, which buyer, and what time window.

Reporting questions

The fastest win. These run against your call log and drill-down reporting without you building a report.

“How many calls came in yesterday, broken down by buyer?”
“Revenue and payout by publisher for the last 7 days, best first.”
“Which campaign had the most converted calls this month?”
“Show me the longest call today and its tags.”
“Of yesterday’s calls to the Auto Insurance campaign, how many were connected, and what was the revenue?”

Then keep going — the assistant can drill into any row you ask about:

Break that top publisher down by state.

Campaigns and routing

“Show me the routing settings for the ‘Auto Insurance’ campaign.”
“Which buyers are in the first priority tier on that campaign, and what are their weights?”
“Pause the ‘Test Campaign’.”
“Disable the ‘Sunrise Insurance’ row in routing without removing it — dry run first.”
“Move ‘Sunrise Insurance’ ahead of the others in the top tier.”

Buyers and caps

Caps are the same ones documented in How to configure Retreaver number caps.

“List my paused buyers.”
“What caps does Sunrise Insurance have right now?”
“Cap Sunrise Insurance at 50 concurrent calls and 200 calls a day.”
“Sunrise hit its hard cap this morning — reset it.”
“Duplicate the Sunrise Insurance buyer and call the copy Sunrise Insurance — Overflow.”
“Set up a new buyer end to end: name Sunrise Insurance, on the Auto Insurance campaign, capped at 50 concurrent calls, business hours 9–5 Eastern weekdays.”

Caller lists

“Is +18885551234 on the Suppressed list for the Auto Insurance campaign?”
“Upload these numbers to the Suppressed list.”
“Have there been any calls to the numbers I uploaded to the Suppressed list this morning?”
“How did last night’s caller list upload go — were there errors?”
“What auto-add rule is on that list?”

Webhooks and integrations

These use the same templates as the webhook configurator.

“What webhooks are on the Auto Insurance campaign, and where do they post?”
“Is there a template for an agent-availability ping? What parameters does it need?”
“Create that webhook on the Sunrise Insurance buyer using our CRM endpoint.”
“Test-fire the CRM webhook and show me exactly what it sent and what came back.”

Housekeeping and audits

Good use of a read-only connection:

“Which buyers have no caps set at all?”
“List every webhook in this company that posts to a URL containing ‘staging’.”
“Which campaigns are paused but still have active numbers pointed at them?”
“Show me every conversion group on the Auto Insurance campaign and what each one pays.”
“Which postback keys are on this company, and where do they point?”

Multi-step recipes

Single prompts get you a long way. The real value is chaining them — each step builds on what the last one returned.

Stand up a buyer and prove the webhook works

“Add a buyer called ‘Sunrise Insurance’ to the ‘Auto Insurance campaign’, cap it at 50 concurrent calls, and show me a dry run first.”
“Apply that. Then create the CRM webhook on that buyer from the configurator template, using our production endpoint.”
“Test-fire that webhook and show me the URL it hit and the response.”
“Put Sunrise Insurance in the top routing tier for California and Oregon — dry run first.”

Diagnose a conversion drop

“Converted calls for the Auto Insurance campaign this week vs last week, broken down by buyer.”
“For the buyer that dropped the most, show me connected vs converted, and the longest converting call.”
“What caps and routing weight does that buyer have right now?”
“Did its hard cap fire this week?”

Audit a company you just switched into

Do this after re-authenticating onto a different company.

“Which Retreaver company am I connected to, and what can I do?”
“List paused campaigns, paused buyers, and webhooks posting to staging URLs.”
“Which buyers have no caps set?”

Getting better answers

Ask which company you’re on first. “Which Retreaver company am I connected to?” One line, and you know exactly which account is about to be touched. This matters most if you manage several — see MCP Authentication.

Ask for a dry run before anything that writes. “…show me a dry run first” returns the exact diff and applies nothing. Especially worth it for routing. (A few actions have no preview — duplicating a buyer, resetting a hard cap, creating or test-firing a webhook.)

Name things exactly. “The Auto Insurance campaign” beats “my main campaign”. If you’re not sure of the exact name, ask it to list them first.

Say the time window. “Yesterday”, “the last 7 days”, “this month” — otherwise reporting questions default to a short recent window.

Ask what changed. After a write, “what exactly did you change?” — most write tools return a diff, so it can tell you precisely.

Start a new session to pick up new tools. Assistants load the tool list when a session starts. When we ship new capabilities, an old session won’t see them.

What it’s not great at yet

Being straight with you:

Bulk edits across many records are slow — the assistant works one call at a time. For “change 200 buyers”, the Browser Platform or the REST API is still the better tool.
RTB reporting isn’t reachable yet. It’s coming with the new RTB Dashboard.
Anything time-critical deserves the Browser Platform for now. See the caveat in Installing MCP Server — this is a young service and we take it down briefly to improve it.
Pixel-perfect exports. The assistant can answer the question and show you the numbers, but a formatted report export still belongs in reporting.