azure-functions-create

Language: Always respond in the same language the user is using.

azure-functions-create — Create Azure Functions App

Guide the user through creating a new Azure Functions project.

Prerequisites

Ensure func (Azure Functions Core Tools v4) is installed. If not, suggest running azure-functions-setup first.

Workflow

Step 1 — Detect Azure MCP tools

Check whether the following Azure MCP tools are available in your current tool list:

• functions_language_list — list supported languages and runtime versions
• functions_project_get — scaffold project-level files
• functions_template_get — list templates (language only) or get a specific template (language + template)

These are provided by the Azure MCP Server (@azure/mcp) and return officially maintained templates across C#, Python, TypeScript, JavaScript, Java and PowerShell.

Also check for the best practices tool:

• get_azure_bestpractices / get_azure_bestpractices_get with resource: azurefunctions

• If available → proceed with Path A (MCP primary).

• If not available → proceed with Path B (composition algorithm fallback).

---

Path A — MCP primary (recommended)

Use the Azure MCP Server as the authoritative source of truth for Azure Functions templates. Do not write function code from scratch when these tools are available.

A.1 Gather requirements & best practices

If get_azure_bestpractices is available, call it first:

Tool: get_azure_bestpractices
resource: azurefunctions
action: code-generation

Apply the returned guidelines (programming models, extension bundles version, authentication levels, project structure, etc.) to every file you generate in the steps below.

Ask the user (or detect from context):

• Language: csharp | python | typescript | javascript | java | powershell
• Trigger / template: let the MCP list decide (step A.3)
• Project name: directory name
• Runtime version (optional): e.g. Node.js 22, Python 3.11, Java 21

A.2 Discover supported languages

Call functions_language_list. Returns supported languages with runtime versions, programming models, and prerequisites. Use this to confirm the user's language choice is supported and to suggest a default runtime version.

A.3 Browse available templates

Call functions_template_get with only the language parameter (omit template). This returns the list of available templates for the chosen language with descriptions. Present the templates to the user and let them pick.

Do not invent or guess template identifiers such as HttpTrigger. Azure MCP template IDs are versioned, language-specific strings returned by the template list. For example, the TypeScript HTTP trigger template is currently returned as http-trigger-typescript-azd.

When the user asks for a common trigger name, map it to one of the template IDs returned by the MCP list before calling the template-get operation. Examples:

User intent	Language	Prefer a returned template ID like
HTTP trigger	typescript	http-trigger-typescript-azd
Timer trigger	typescript	timer-trigger-typescript-azd
Blob trigger	typescript	blob-eventgrid-trigger-typescript-azd
Queue / Service Bus trigger	typescript	servicebus-trigger-typescript-azd

If a template-get call fails with "template not found", immediately recover by calling functions_template_get again with only language, then select the closest returned template ID instead of retrying the failed alias.

A.4 Initialize the project

Call functions_project_get:

Tool: functions_project_get
language: <chosen language, e.g. typescript>

Returns project-level files (host.json, local.settings.json, package.json / requirements.txt / pom.xml / .csproj, tsconfig.json, etc.). Write these into the target directory.

A.5 Add the function

Call functions_template_get with both language and template:

Tool: functions_template_get
language: <chosen language, e.g. typescript>
template: <chosen returned template ID, e.g. http-trigger-typescript-azd>
runtime-version: <optional, e.g. 22>
output: <optional, "New" (default) or "Add" for existing projects>

Returns the full function source code plus any required app settings and additional package dependencies. Write the returned file(s) into the project and merge any extra settings into local.settings.json and any extra packages into the dependency manifest.

⚠️ Large template output: Some templates (especially *-azd variants that include infrastructure files) can produce very large output (100KB+). If the tool output is truncated or saved to a temporary file, read the file, parse the JSON files array, and write each file individually. After writing files, run npm install (or the equivalent package manager command) to generate lock files rather than relying on lock files from the template output.

A.6 Verify

For TypeScript and other compiled-language projects, build first:

npm run build   # TypeScript / JavaScript
# dotnet build  # C#
# mvn package   # Java

Then perform an end-to-end local verification, not just a host start:

func start

After the host reports the function endpoints/listeners:

• HTTP triggers: send an actual request to the local endpoint and verify the status code and response body, for example curl http://localhost:7071/api/<FunctionName>?name=World.
• Timer triggers: verify the listener starts and, when practical, temporarily use a short development-only schedule or manual invocation approach; restore the user's intended schedule before finishing.
• Storage, Cosmos DB, SQL, Redis, Dapr, or other service-backed triggers/bindings: load azure-functions-common/references/local-emulators.md, identify the required local emulator or development service, and run a realistic message/blob/document/event through the trigger when the user wants E2E verification.
• Before installing or starting any emulator/local service: ask the user for confirmation. If the user says the emulator is not needed, unavailable, or should be skipped, do not install it; record that emulator-backed E2E was skipped and provide manual/Azure test steps instead.
• When no practical local emulator exists: explain the limitation, suggest a temporary Azure dev resource or deployment-based test, and keep the local verification to build + host/listener startup.

---

Path B — Composition algorithm fallback

Use this path only when the Azure MCP tools are not available. When falling back, show this notice to the user verbatim (translate to the user's language if needed):

ℹ️ Azure MCP tools were not found; using the manifest-based fallback path. Enabling the Azure MCP Server unlocks dynamic template discovery and composition. Run azure-functions-setup to configure it.

B.1 Fallback algorithm

Follow this manifest-based fallback algorithm:

1. FETCH MANIFEST
   GET https://cdn.functions.azure.com/public/templates-manifest/manifest.json
   If fetch fails → fall back to:
     https://github.com/Azure/azure-functions-templates/blob/dev/Functions.Templates/Template-Manifest/manifest.json
   If both fail → fall back to known-good Azure-Samples/functions-quickstart-* repos
   If all fail → report error and ask user to retry later

2. FILTER TEMPLATES
   Filter by: language, resource, iac

3. CHECK SINGLE-TEMPLATE MATCH
   If one template covers ALL requirements → use it alone

4. SELECT TEMPLATES
   - Trigger template (REQUIRED) — base project with IaC
   - Binding templates (OPTIONAL) — extract patterns only

5. DOWNLOAD TEMPLATES
   For each template:
   - If folderPath == "." → ZIP download + unzip
   - If folderPath != "." → fetch tree + raw github url file downloads
   - Fallback: git clone --depth 1

6. COMPOSE
   - Use trigger template as BASE
   - EXTRACT binding patterns from binding templates
   - MERGE IaC resources, RBAC roles and settings
   - ADD user's custom business logic

7. TRIM unused demo code (keep AzureWebJobsStorage)

8. WRITE all files

9. DEPLOY: azd up --no-prompt

B.2 Quick code reference

For minimal HTTP trigger snippets per language (last-resort fallback when the manifest is also unavailable), see references/language-snippets.md.

B.3 Verify

Build compiled projects first, then perform the same local E2E verification standard used in Path A:

func start

• For HTTP triggers, send a real request to the local endpoint and validate the response.
• For non-HTTP triggers, consult azure-functions-common/references/local-emulators.md and use an emulator/local service when practical.
• Ask before installing or starting emulators. If the user declines, skip emulator-backed E2E and document the skipped verification plus manual/Azure test steps.

---

Adding functions to existing projects

If host.json already exists, do not re-initialize. Instead:

• MCP path: call functions_template_get with the same language as the existing project and specify the desired template name. Write the returned file.
• Fallback path: fetch the manifest, filter for the desired template by language and resource, download the template source, and merge the function files into the existing project.

After Creation

✅ Your project is scaffolded! Next: use azure-functions-deploy to deploy to Azure.

Next steps

• On success, suggest azure-functions-deploy because the project is ready to deploy to Azure.
• On failure, suggest azure-functions-setup to verify local tooling and prerequisites.