AI agents are embedded in the software development life cycle, from code generation and debugging to incident triage. These agents operate in isolation from observability data to help engineers debug issues and respond to incidents, which results in constant context switching. Engineers start a debugging session, realize they need a trace or an error log, and have to switch workspaces, find the data, and rebuild context before they can continue.
To close that gap, Datadog now offers integrations for every major AI chat and agent, including Claude Code, Claude Desktop, Claude Cowork, ChatGPT, Codex, OpenCode, and Cursor. Powered by the Datadog MCP Server, the integrations give AI agents a more secure, structured interface to query Datadog’s observability data and tools.
How Datadog’s MCP integration works
All integrations connect through the Datadog MCP Server, which exposes Datadog’s observability capabilities as tools that AI agents can call through natural language, from log search and metric queries to incident lookup, service governance, and more. The server authenticates through OAuth and uses HTTP transport, so it works reliably across local and remote agent environments.
Integration and setup vary by platform type:
