Datadog unveils MCP Server for governed AI observability

Datadog has released an MCP Server that gives AI agents controlled access to real-time observability data, including logs, metrics, and traces.

The launch targets software teams starting to use AI agents in live development and operational workflows. As this shift accelerates, teams are asking how agents can read production data and take action without bypassing governance controls.

Engineers rely on observability data to diagnose incidents, understand system performance, and track production changes. As AI tools move from generating code to interacting with live systems, access to this data has become a growing focus for security and compliance teams.

Agent access

Datadog's MCP Server is an interface for agentic systems, connecting AI workflows to Datadog's observability platform while keeping access within existing security, governance, and audit controls.

The product is aimed at organisations embedding AI agents into engineering workflows across development and operations. It positions Datadog as a data source for agents asked to investigate alerts, review telemetry, and propose or carry out remediation steps.

"Datadog delivers AI solutions that transform complexity into clarity and blind spots into security, helping protect global businesses and make operations seamless. We are always listening to our customers and the biggest problems they are facing in their day-to-day work, which is why we are excited to launch our MCP Server as the latest way to help teams become more efficient in using Datadog to build and scale AI systems across their organisations," said Yanbing Li, Chief Product Officer at Datadog.

Li described the product as part of a move toward AI systems working directly with production environments. "By combining telemetry from Datadog's unified observability platform into teams' AI workflows, we are enabling the next stage of AI-native development-moving from simply AI copilots to AI operating on live production systems," Li said.

Debugging workflows

The MCP Server can feed live logs, metrics, and traces into AI coding agents and integrated development environments during incident investigations. Datadog cited Codex, Claude Code, and Cursor as examples of tools that can be used in these workflows.

This approach reflects an emerging pattern in software operations, where incident response starts in chat-style interfaces and agent-driven tooling. Engineers want to query telemetry, check recent deployments, and view service dependencies without switching between multiple dashboards and consoles.

Datadog also positioned the MCP Server as a way for custom AI agents to access "real-time observability and intelligence." It said the server exposes Datadog signals used for detection and remediation, allowing agents to investigate issues and respond automatically under defined controls.

Governance focus

As automation increases in production operations, companies are expanding governance frameworks. These typically include tighter permissioning, audit logging, and approval steps for actions that can affect customer-facing systems. AI agents add complexity because they can combine broad system access with autonomous decision-making.

Datadog framed the MCP Server as a way to keep AI operations governed when agents interact with production telemetry and workflows. It also said the product reduces integration overhead and supports compliance expectations for teams operationalising AI agents.

A technical element of the launch is a protocol-based approach to agent communication. Datadog described the MCP Server as a "dynamic, purpose-built protocol" designed to reduce the risk of breaking changes in AI workflows that depend on telemetry access.

That risk is becoming more visible as engineering teams connect AI tooling to internal APIs and operational systems. Changes to schemas, query formats, or authentication layers can disrupt automations embedded in day-to-day incident response routines.

Product availability

The MCP Server is generally available, though Datadog did not disclose pricing.

The launch also reinforces Datadog's broader message that observability is becoming a core data layer for AI-assisted operations. Telemetry provides a record of what systems are doing in real time and over time, informing root-cause analysis, performance tuning, and security investigations.

"AI compounds complexity, especially with its pace of innovation. Datadog is helping to solve that complexity for customers with launches like MCP Server by enabling autonomy across Dev, Ops and Security teams so that they can not only detect, decide and act on issues within Datadog, but also build, deliver and evaluate software throughout the development process," said Li.

Datadog said the MCP Server is intended for teams embedding AI agents into development and operational workflows that want access to live telemetry within existing security and governance controls.