The MCP connector that lets your AI make real phone calls — book the table, chase the refund, cancel the gym membership — and report back with a full transcript. Every call dials out from your own number.
Your AI isn't literally on the phone — it's the dispatcher. It hands our real-time voice agent an objective; the agent dials, listens, speaks and negotiates with the human on the other end; your AI polls the live transcript and reports back.
Add mcp.getdylr.com/mcp to Claude, ChatGPT or Grok. One URL, standard MCP OAuth — about 60 seconds.
We call you with a code. From then on, every call your AI makes shows your caller ID — because it's calling for you, as you.
"Call the dentist and move my Thursday cleaning to next week." Your AI dispatches it and keeps you posted.
Live transcript while it talks. When it hangs up: outcome grade + summary with every price, time and confirmation number.
dylr MCP speaks MCP — the open standard all three major AI apps support. Choose yours:
If your team runs on dylr, you don't sign up again and you don't pay again. Sign into the dashboard with your existing dylr login and the bundle switches on:
Calls can only dial out from numbers you own — your verified cell or your org's business lines. No spoofing, no burner pools. Businesses can call you right back.
Every call opens with an honest introduction: "I'm an AI assistant calling on behalf of…" — non-negotiable, baked into the agent.
Emergency and premium numbers are unreachable. Per-day and per-destination caps. Hard time limits. Cancel any call mid-flight.
If someone says "don't call me again," the agent apologizes, hangs up, and that number goes on your do-not-call list automatically and permanently. Not in 10 business days. Immediately.
No — and that's by design. Claude (or ChatGPT/Grok) is the dispatcher: it sends our platform an objective over MCP. A dedicated real-time voice agent — speech recognition + a fast LLM + neural text-to-speech running on our edge — has the actual conversation, turn by turn, in about a second per reply. Your AI polls the live transcript and gives you the outcome. Think of it as Claude delegating to a very capable intern with a headset.
Yes — for what it's built for. You're placing calls from your own number, for your own errands, with an agent that discloses it's an AI. What's not allowed (and blocked or banned): telemarketing blasts, robocall spam, impersonation, and calling emergency services. See the Terms.
Your phone number. That's the whole point — the pizzeria sees your caller ID and can call you back on it. The agent also gives them your number if they ask how to reach you.
No. dylr MCP is included with dylr. Sign into the dashboard with your existing dylr credentials — your org's active phone numbers import as caller IDs automatically and you get Pro limits at no charge. Buy a new number in dylr and your AI can call from it.
It detects the machine, waits for the beep, and leaves a short, useful message with your callback number. The transcript shows exactly what it said.
Yep — it waits politely. Calls have a hard time cap (5 min free / 10 min Pro), so a brutal hold queue ends the call rather than running forever.
It stops anyone from spoofing calls from numbers they don't own, and it means every call is traceable to a real, reachable person — you. It's the anti-abuse design and the feature.