How a six-client outbound agency replaced per-row AI credit markups by connecting Claude Desktop to Sorzio MCP with an encrypted Anthropic BYOK key, lifting positive reply rates 3.4x.
Meridian manages outbound sequences for six B2B retainers, shipping roughly 48,000 net-new prospect lines a month. Bundled credit platforms charged a steep inference margin on every row and still produced hallucinated product claims when websites were thin.
Strategists spent 11 hours every week exporting CSVs between scrapers, chat windows, and sequencer sheets to spot-check tone before launch.
Connecting Claude to https://www.sorzio.com/api/mcp let each strategist upload a Hosted List, preview 10 rows in chat, tighten the prompt once, and run the full list with self-healing output verification enabled.
With Meridian's own API key stored in AES-256-GCM encrypted BYOK settings on Scale ($99/month, 3 seats included), Sorzio billed $0.00 on all 48,000 generative calls.
create_list + preview_prompt inside Claude via OAuth 2.1 MCP
Strategists inspected column headers and 10 live rows in chat before approving full execution.
run_scrape Auto bundle (homepage + 5 high-signal subpages)
Grounded every row in real case studies, pricing tiers, and open roles before writing a single word.
run_ai with self-healing verification on agency API key
Provider tokens billed directly to Meridian's provider account at list rate; unsupported claims revised automatically.
“We test 10 rows in Claude, read the exact sentences, and approve 8,000 rows without touching a CSV. Zero markup on our own API key saved more in week one than our Scale seat costs all year.”