DylrDev connects Claude, ChatGPT, Cursor, and any MCP client to Twilio, Five9, and Genesys Cloud. Reads just run. Dangerous actions wait for a person. Every call is logged.
Add a Twilio, Five9, or Genesys Cloud account. We check the credentials with one read call, then keep them encrypted. You get an MCP URL for that account alone.
Stopping a campaign or deleting a queue needs an owner or admin to approve that exact call, or to arm the connection for 30 minutes. The AI gets a link to send you.
Every tool call and every change: who, through which app, and how it was cleared. Never the arguments.
claude mcp add --transport http dylr https://dylr.dev/mcp/<connection> --header "x-api-key: dylr_…"
Fixed tools that say what they do. Never a raw API call.
Phone numbers, Studio flows, TaskRouter workspaces, queues, and workers. Create task queues; deleting one is dangerous.
Campaigns, lists and contacts, users and skills, dispositions, prompts, IVR scripts, reports, and live stats. Deletes, user changes, and changes to anything a running campaign uses are dangerous.
Queues, users and skills, schedules, outbound campaigns and lists, Architect flows, virtual agents, and knowledge. Adding contacts a live campaign could dial, publishing over a live flow, and test chats are dangerous.
It takes a few minutes: an account, the platform's credentials, and one URL.