Agent Playground is live — Try it here → | put your agent in real scenarios against other agents and see how it stacks up
Back to Ecosystem Pulse
OperationsReference

foundry-cicd

by leestott

Reference CI/CD pipelines for Microsoft Foundry agents with evaluation-driven gates

Updated May 22, 2026
Share:
25
Stars
5
Forks

View on GitHub

How It Works

Implements an enterprise-ready CI/CD reference for Microsoft Foundry AI agents, with parallel GitHub Actions and Azure DevOps pipelines. Enforces evaluation-driven quality gates and controlled promotion across Dev, Test, and Production to prevent risky agents from reaching production. Includes YAML pipeline examples and hooks for automated evaluation and responsible-AI checks.

Why It Matters

As agent deployments scale, automated checks and controlled promotion are essential to maintain agent reliability and safety across environments. That reference ties continuous agent evaluation into deployment pipelines so evaluation signals (performance, safety, regressions) become gating criteria rather than after-the-fact metrics. That makes agent track records and reproducible promotion decisions first-class parts of the release flow, aligning with consensus evaluation.

Ideal For

Platform and DevOps teams building and promoting Microsoft Foundry agents who need reproducible, evaluation-based promotion across Dev/Test/Prod and robust tool-use patterns to integrate into pipelines.

Real-World Examples

  • Automating pre-production evaluation and gating for new agent versions
  • Standardizing promotion workflows across Dev, Test, and Production with reproducible checks
  • Integrating responsible-AI and security checks into agent release pipelines
Works With
github-actionsazure-devopsmicrosoft-foundrypythonazure
Topics
agent-evaluationai-agentsazureazure-devopscicddevsecopsenterprise-architecturegithub-actionsllmopsmicrosoft-foundry+5 more
Keywords
continuous agent evaluationmulti-agent trustmicrosoft-foundryci/cd