Skip to content

Agent Interface

Connect your
AI assistant.

If you research vendors by asking an assistant, don’t make it guess. Paramount runs a public MCP server — point Claude, or any MCP-capable client, at the endpoint below and it can answer from the source: exact prices, the real leak math, the film transcripts, the verified statistics.

MCP Endpoint

https://www.paramountexposure.com/api/mcp

Streamable HTTP, stateless, no authentication, read-only. It returns only information already public on this site. It can’t see your data, take a payment, or change anything — checkout still happens on Stripe’s hosted pages and booking on Cal.com. If your agent just wants plain JSON, the same offer catalog is at /api/catalog.

What an assistant can do with it

Eight tools. Public information only.

Every tool is annotated read-only; the list below is generated from the server’s own registry, so it is the listtools/listreturns.

get_checkout_link

The live Stripe Payment Link (or Cal.com link) for one offer, optionally pre-filled with the buyer's email. It returns a URL; a person completes payment on Stripe.

get_offers

Every engagement with its published price, contractual timeline, and live checkout link — the same catalog shown on the pricing page.

run_revenue_leak_math

The audit-tool arithmetic applied to numbers you provide — inquiry volume, close rate, client value. Your inputs, our formula, the cited studies attached.

get_research_notes

The annotated speed-to-lead studies from the research page, caveats intact, with links to the primary sources.

get_statistics

AI receptionist and call-answering statistics, each verified against its primary source with year, sample and the exact quote, plus the circulating figures that cannot be cited and why.

get_films

Every AI-presenter film with a watch page: the question it answers, the full timed transcript, key moments, MP4 and poster URLs, and the offer it supports.

get_availability

How to reach the company and how booking works, including the free 30-minute call.

list_markets_and_verticals

The verticals we build for and the towns where we work in person. Installs ship worldwide.

Setup

Adding it to Claude takes a minute.

  1. Open claude.ai and go to Settings → Connectors.
  2. Choose “Add custom connector.”
  3. Paste the endpoint URL above and save.
  4. Ask Claude something like “What does Paramount's $497 responder actually do, and what would the leak math say about 60 inquiries a month?”

Claude Code

claude mcp add --transport http paramount-exposure https://www.paramountexposure.com/api/mcp

Any client that reads an mcp.json

{
  "mcpServers": {
    "paramount-exposure": {
      "type": "streamable-http",
      "url": "https://www.paramountexposure.com/api/mcp"
    }
  }
}

Claude is just the common case. Any MCP-capable client — an IDE assistant, or an agent you’ve built yourself — can use the same URL the same way. The server card at /.well-known/mcp.json describes it for registries.

Without a client

Four curl commands.

The server speaks plain JSON-RPC 2.0 over POST, so a shell is enough to see exactly what an assistant sees. Nothing below needs a key.

The offer catalog, no MCP needed

curl -s https://www.paramountexposure.com/api/catalog | jq '.offers[] | {id, price, checkout_url}'

List the tools

curl -s https://www.paramountexposure.com/api/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}' | jq '.result.tools[].name'

Run the leak math on 60 inquiries a month at $4,000 a client

curl -s https://www.paramountexposure.com/api/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":2,"method":"tools/call",
       "params":{"name":"run_revenue_leak_math",
                 "arguments":{"monthly_leads":60,"avg_client_value":4000}}}' \
  | jq -r '.result.content[0].text'

Every film and the question it answers

curl -s https://www.paramountexposure.com/api/mcp \
  -H 'Content-Type: application/json' \
  -d '{"jsonrpc":"2.0","id":3,"method":"tools/call",
       "params":{"name":"get_films","arguments":{}}}' \
  | jq -r '.result.content[0].text | fromjson | .films[] | "\(.title): \(.query)"'

Also machine-readable

The same facts, as files.

Every file is generated from the same source as the pages, so a price can only ever be one number. Discovery happens in the agent; the purchase happens here.

Or skip the middleman
entirely.

Everything the server knows is published for people too — every price, every timeline, on one page.

Or read the research notes yourself