Logic App AI Agent

template to deploy logic app workflows as an ai agent

Logic App Standard is Microsoft's low code offering for implementing enterprise integrations. It offers Connectors which can save you time from building everything yourself.

Azure Logic Apps can now integrate AI agents, which use large language models (LLMs) and an "agent loop" to perform complex, multi-step tasks autonomously. These agents can reason, make decisions, and use tools (prebuilt Logic App actions) to complete a workflow that evolves based on results. This is a new approach that complements existing structured automation with an AI-first, adaptive pattern, allowing agents to plan, act, learn, and reason within the low-code platform.

This template includes a Logic App Standard and AI Foundry deployment.

Application architecture

This template utilizes the following Azure resources:

• AI Foundry for for building, deploying, and managing AI and generative AI applications
• Azure Logic App Standard to design the workflows
• Azure Key Vault for securing secrets
• Azure Monitor for monitoring and logging

How to get started

1. Install Visual Studio Code with Azure Logic Apps (Standard) extensions
2. Create a new folder and switch to it in the Terminal tab
3. Run azd auth login
4. Run azd init -t https://github.com/marnixcox/logicapp-ai-agent

Now the magic happens. The template contents will be downloaded into your project folder. This will be the next starting point for building your AI agents.

Contents

The following folder structure is created.

├── infra                      [ Infrastructure As Code files ]
│   ├── aifoundry              [ AI Foundry files ]
│   ├── keyvault               [ Key Vault files ]
│   ├── logicapp               [ Logic app files ]
│   ├── storage                [ Storage files ]
│   ├── main.bicep             [ Main infrastructure file ]
│   └── main.parameters.json   [ Parameters file ]
├── src                        [ Application code ]
│   └── workflows              [ Azure Logic App Standard ]
└── azure.yaml                 [ Describes the app and type of Azure resources ]

Key Infrastructure Files

• aifoundry/aifoundry.bicep - AI Foundry
• aifoundry/role-assignment.bicep - To give Logic Apps access to AI Foundry
• keyvault/keyvault.bicep - Key Vault
• logicapp/plan.bicep - Logic App Plan
• logicapp/workflows.bicep - Logic App Standard
• storage/storage.bicep - Storage account for Logic App
• main.bicep - Main orchestration
• monitoring.bicep - Observability stack
• resources.bicep - Resource collection

Provision Infrastructure

Let's first provision the infra components. Add a resourceToken parameter to the main.parameters.json to set the application name or leave it out to have a random name generated.

 "resourceToken": {
      "value": "appname"
    }

• Run azd provision

Environment

First time an environment name, subscription and location need to be selected. These will then be stored in the .azure folder.

Allowed values for environment are dev tst acc prd

Resource group and all components will be created.

Deploy Contents

After creating Azure Logic App Standard workflows these can be deployed with another single command.

• Run azd deploy

Workflows will be deployed into the existing infra components.

Connections

In order to setup the AI Foundry Connections in the connections.json file the following parameters are available in the Logic App Standard instance:

AI_FOUNDRY_NAME AI_FOUNDRY_ENDPOINT AI_PROJECT_NAME AI_PROJECT_ENDPOINT