Delight.ai has launched Delight Agent MCP, a Model Context Protocol connection that lets customer support teams query live data from their Delight.ai workspace through external AI assistants such as Claude, Codex, and Cursor. Support managers can now ask questions like "Why did customer satisfaction drop this week?" or "How's my agent doing today?" and get answers backed by operational data, without opening a dashboard or exporting a report.
The new server works in the opposite direction of the MCP Tool Import feature Delight.ai introduced earlier this year. Tool Import lets customers bring their own MCP servers into Delight.ai agents. Delight Agent MCP publishes Delight.ai's own server, so external AI clients can securely access and query live workspace data.
Model Context Protocol is the open standard that lets AI assistants connect to external data sources and tools. Support professionals who want to understand how these connections work can explore MCP Courses.
The launch fits a broader industry shift toward AI assistants that interact directly with enterprise applications. Users can retrieve information and complete tasks in natural language instead of moving between multiple interfaces.
"Enterprise software is becoming something you talk to, not something you log into," said John Kim, CEO and co-founder of Delight.ai, a Sendbird company. "Instead of opening dashboards, exporting data and piecing together answers, teams should be able to ask a question and receive an answer grounded in operational data. That's where AI-powered work is headed."
What Delight Agent MCP lets teams do
At launch, Delight Agent MCP gives organizations direct access to more than 35 operational tools. The capabilities fall into four areas:
- Agent performance: Check resolution rate, CSAT, and category breakdowns.
- Conversation search: Search and read full conversation transcripts.
- Quality monitoring: Identify low-confidence responses, safeguard flags, and user feedback.
- Write operations: Create and update actionbooks, tools, knowledge sources, and agent configuration.
The tool was developed through Delight.ai's own internal workflows before becoming generally available. Customer demand for direct, standards-based access to live operational data without export workflows helped shape the final product.
Why this matters for customer support teams
The immediate benefit is speed. A support manager who wants to know why CSAT dipped no longer has to request an export, wait on a report, or dig through a dashboard. They can ask an AI assistant and get an answer pulled from live conversation data.
The write operations are the bigger change. Because the MCP server can update actionbooks, tools, knowledge sources, and agent configuration, it moves AI assistants from read-only reporting into operational work. Teams will want to decide which of these actions they're comfortable authorizing, and what review process to put around them.
Support leaders who want to plan for this shift can learn more through AI for Customer Support resources.
Your membership also unlocks:
