What is Agent Landing Zone?
Agent Landing Zone is a secure, enterprise-ready landing zone from Microsoft's Azure organisation for deploying AI agent applications on Microsoft Foundry. It sits in the RAG Frameworks & Platforms category of the Retrieval & Knowledge (RAG) tools and is open-source, written in PowerShell and licensed MIT. To get started: clone the repository.
Overview
Agent Landing Zone is a secure, enterprise-ready landing zone from Microsoft's Azure organisation for deploying AI agent applications on Microsoft Foundry. It is the former GPT-RAG Solution Accelerator: with the v4.0.0 release on 2026-10-03 the repository moved from Azure/GPT-RAG to Azure/agent-landing-zone, because the project had grown from a RAG accelerator into a foundation for agent applications in general. It is not a library you import but an Azure Developer CLI template: you run azd from the repository and get the infrastructure, and optionally a working application, provisioned in your own subscription.
The default application is a trio of services, renamed in v4.0.0 to agent-app-ui (the chat front end), agent-app-orchestrator (the agentic workflow, built on the Microsoft Agent Framework) and agent-app-ingestion (which extracts, chunks and indexes your content), with versions pinned in the repository's manifest.json. Foundry IQ is the default retrieval backend, and the orchestrator can run either as a Container App or as a hosted agent in Foundry Agent Service. You can also swap the trio out entirely: describe your own application in an app-definition.json file and the landing zone deploys that instead.

Infrastructure and application are deliberately separate layers. azd provision builds the foundation (network, Foundry project, search, storage, databases, identities and Key Vault) defined in Bicep and incorporated from the AI Landing Zone v2.7.3 release, and publishes its outputs to App Configuration so applications can discover it; azd deploy adds the application; azd up does both. The architecture is Zero-Trust, with services behind private endpoints in an isolated network, and it is MIT-licensed. v4.0.0 supports new deployments only: existing GPT-RAG environments are not upgraded in place and must be redeployed, while GPT-RAG release tags stay available for environments that keep running on them.
What it does
- Infrastructure-only or full-stack deployment with the native azd verbs: azd provision for the landing zone, azd deploy for the application, azd up for both
- A default agentic RAG application of three components, agent-app-ui, agent-app-orchestrator and agent-app-ingestion, pinned per release in manifest.json
- Bring your own application through an app-definition.json file, with Container Apps and Foundry hosted-agent samples in samples/custom-app
- Foundry IQ as the default retrieval backend, with orchestration on the Microsoft Agent Framework as a Container App or a Foundry hosted agent
- Zero-Trust architecture with private endpoints, network isolation and managed identities, all defined as Bicep infrastructure as code
- Platform outputs (Foundry project, registry, search, storage, Cosmos DB, identities, network mode) published to App Configuration for applications to read
Getting started
You need an Azure subscription, the Azure Developer CLI and Git; full prerequisites, network-isolated deployment and configuration are covered in the Agent Landing Zone section of the AI Landing Zones documentation. The repository's v4.0.0 release was validated with a fresh basic deployment without network isolation, which is what the steps below set up. Run azd from the repository root.
Clone the repository
azd runs from the Agent Landing Zone repository root.
git clone https://github.com/Azure/agent-landing-zone.git
cd agent-landing-zoneSign in and choose a basic environment
Authenticate the Azure Developer CLI, then turn off network isolation for a demo environment. Leave it on for a Zero-Trust production deployment, following the documentation's network-isolated guide.
azd auth login
azd env set NETWORK_ISOLATION falseDeploy the full stack
One command provisions the landing zone infrastructure and deploys the UI, orchestrator and ingestion applications.
azd upOr deploy the infrastructure first
To stand up only the landing zone now and add the application later, split the two steps. A repeated azd provision keeps images that are already published.
azd provision
# later, when you are ready to ship the application
azd deployDeploy your own application
Copy a folder from samples/custom-app into your own repository, validate its app-definition.json, then point a new azd environment at it. One environment hosts exactly one application.
python -m config.appdefinition --validate <folder>
azd env set AGENTLZ_APP_DEFINITION <folder>
azd upCommands and code are distilled from the project's own documentation — always check the official repo for the latest.
When to use it
- Reach for it when an enterprise on Azure needs a chat assistant grounded in internal documents and data, and wants a reference deployment rather than building one from parts
- Reach for it when a platform team wants to provision a secure Foundry foundation once and let application teams deploy their own agents onto it
- Reach for it when security review demands private endpoints, network isolation and least-privilege access from day one
- Reach for it when you ran GPT-RAG and are planning a new environment under the renamed project, since v4.0.0 is a redeploy rather than an in-place upgrade
How Agent Landing Zone compares
Agent Landing Zone alongside other open-source rag frameworks & platforms tools AI/TLDR tracks, ranked by GitHub stars.
| Tool | Stars | What it does |
|---|---|---|
| Dify | ★ 162k | An open-source platform with a visual workflow builder for creating LLM and RAG applications without writing much code. |
| graphify | ★ 151k | Turns a folder of code, docs, PDFs and images into a local knowledge graph with tree-sitter AST parsing and Leiden communities — queryable by agents over MCP, no vector store. |
| RAGFlow | ★ 94.4k | A RAG engine built around deep document understanding that turns complex files into a grounded, citation-backed question-answering layer. |
| Context7 | ★ 65.1k | Context7 pulls current, version-specific documentation and code examples for any library and feeds them into your LLM, available as a CLI skill or an MCP server. |
| Pathway | ★ 62k | A Python framework with a Rust streaming engine that keeps ETL, real-time analytics and RAG pipelines continuously up to date as source data changes. |
| LightRAG | ★ 40.8k | A graph-based RAG system that builds an entity-and-relationship knowledge graph for fast retrieval and easy incremental updates. |
| PageIndex | ★ 40.5k | PageIndex turns long PDFs into a table-of-contents tree and uses LLM reasoning to retrieve relevant sections, with no vector database and no chunking. |
| Agent Landing Zone | ★ 1.2k | Microsoft's Zero-Trust landing zone for deploying AI agent applications on Microsoft Foundry, formerly the GPT-RAG accelerator |
