Getting Started with Software Delivery MCP Tools
This product is not supported for your selected
Datadog site. (
).
このページは日本語には対応しておりません。随時翻訳に取り組んでいます。
翻訳に関してご質問やご意見ございましたら、
お気軽にご連絡ください。
Overview
The Datadog MCP Server enables AI agents to access your Software Delivery data through the Model Context Protocol (MCP). The software-delivery toolset provides tools for interacting with CI Visibility and Test Optimization directly from AI-powered clients like Cursor, Claude Code, or OpenAI Codex.
Use cases
The Software Delivery MCP tools unlock AI-assisted workflows for:
- Debugging pipeline failures: Ask your AI agent to find recent pipeline failures, analyze error logs, and suggest fixes.
- Identifying flaky tests: Query for flaky tests in your repository and get prioritized recommendations for which to fix first.
- Analyzing CI performance: Get aggregated statistics on pipeline durations, failure rates, and trends over time.
- Triaging test failures: Understand which tests are failing, their ownership, and historical patterns.
The software-delivery toolset includes the following tools:
search_datadog_ci_pipeline_events- Search CI pipeline events with filters and get details on failures, durations, and statuses.
aggregate_datadog_ci_pipeline_events- Aggregate CI pipeline events for statistics like average durations, failure counts, and percentile analysis.
get_datadog_flaky_tests- Find flaky tests with triage details including failure rate, category, owners, and CI impact.
aggregate_datadog_test_events- Aggregate test events to analyze reliability and performance trends.
search_datadog_test_events- Search test events with filters and get details on them.
Example prompts
After you are connected, try prompts like:
- Show me all the pipelines for my commit
58b1488. - What’s causing the
integration-test job to fail on my branch? - Find active flaky tests for the checkout service sorted by failure rate.
- How many failed pipelines have there been in the last 2 weeks?
- Show me the 95th percentile of pipeline duration grouped by pipeline name.
Setup
To use the Software Delivery tools, connect to the Datadog MCP Server with the software-delivery toolset enabled. Add the toolsets query parameter to the endpoint URL for your Datadog site:
https://mcp./api/unstable/mcp-server/mcp?toolsets=core,software-delivery
For full setup instructions including client configuration for Cursor, Claude Code, VS Code, and other AI clients, see Set Up the Datadog MCP Server.
Further reading