Rediger

Azure MCP Server tools for Microsoft Foundry Extensions

The Azure MCP Server lets you manage Microsoft Foundry Extensions resources, including creating chat and text completions, generating embeddings, listing models, and working with knowledge indexes, with natural language prompts.

Microsoft Foundry is a platform for deploying and managing custom AI models in Azure. It provides tools and services for training, fine-tuning, deploying, and monitoring AI models in production environments.

Knowledge index commands require the endpoint of your Microsoft Foundry project. OpenAI commands require the resource name and resource group. The resource get command can list resources without providing a resource name.

Note

Tool parameters: The Azure MCP Server tools define parameters for data they need to complete tasks. Some of these parameters are specific to each tool and are documented here. Other parameters are global and shared by all tools. For more information, see Tool parameters.

Knowledge: List knowledge indexes

Get a list of knowledge indexes from Foundry:

  • Find knowledge indexes created within Foundry projects.
  • Use these indexes with AI agents for knowledge retrieval and RAG applications.
  • The list updates as you create new indexes or update existing ones.

Example CLI command

azmcp foundryextensions knowledge index list \
  --endpoint <endpoint>
Parameter Type Required Description
endpoint string Yes The endpoint URL for the Microsoft Foundry project or service. The endpoint follows this pattern: https://<foundry-resource-name>.services.ai.azure.com/api/projects/<project-name>.

Tool annotation hints:

Destructive Idempotent Open World Read Only Secret Local Required

Knowledge: Get index schema

Get the detailed schema configuration of a specific knowledge index from Foundry.

This operation shows you comprehensive information about the structure and configuration of a knowledge index, including field definitions, data types, searchable attributes, and other schema properties. Use this schema information to understand how the index structures and indexes your data for searching.

Example CLI command

azmcp foundryextensions knowledge index schema \
  --index <index> \
  --endpoint <endpoint>
Parameter Type Required Description
index string Yes The name of the knowledge index.
endpoint string Yes The endpoint URL for the Microsoft Foundry project or service. The endpoint follows this pattern: https://<foundry-resource-name>.services.ai.azure.com/api/projects/<project-name>.

Tool annotation hints:

Destructive Idempotent Open World Read Only Secret Local Required

OpenAI: Create chat completions

Create chat completions by using Azure OpenAI in Foundry. Send messages to Azure OpenAI chat models deployed in your Foundry resource and receive AI-generated conversational responses. Supports multi-turn conversations with message history, system instructions, and response customization.

Example CLI command

azmcp foundryextensions openai chat-completions-create \
  --resource-name <resource-name> \
  --deployment <deployment> \
  --message-array <message-array> \
  --resource-group <resource-group> \
  [--max-tokens <max-tokens>] \
  [--temperature <temperature>] \
  [--top-p <top-p>] \
  [--frequency-penalty <frequency-penalty>] \
  [--presence-penalty <presence-penalty>] \
  [--stop <stop>] \
  [--stream <stream>] \
  [--seed <seed>] \
  [--user <user>]
Parameter Type Required Description
resource-name string Yes The name of the Microsoft Foundry resource.
deployment string Yes The name of the deployment.
message-array string Yes JSON array of messages in the conversation. Each message should have role and content properties.
resource-group string Yes The Azure resource group name.
max-tokens string No The maximum number of tokens to generate in the completion.
temperature string No Controls randomness in the output. Lower values make it more deterministic.
top-p string No Controls diversity via nucleus sampling (0.0 to 1.0). Default is 1.0.
frequency-penalty string No Penalizes new tokens based on their frequency (-2.0 to 2.0). Default is 0.
presence-penalty string No Penalizes new tokens based on presence (-2.0 to 2.0). Default is 0.
stop string No Up to 4 sequences where the API stops generating further tokens.
stream string No Whether to stream back partial progress. Default is false.
seed string No If specified, the system makes a best effort to sample deterministically.
user string No User identifier for tracking and abuse monitoring.

Tool annotation hints:

Destructive Idempotent Open World Read Only Secret Local Required

OpenAI: Create embeddings

Create embeddings using Azure OpenAI in Foundry. Generate vector embeddings from text using Azure OpenAI deployments in your Foundry resource for semantic search, similarity comparisons, clustering, or machine learning.

Example CLI command

azmcp foundryextensions openai embeddings-create \
  --resource-name <resource-name> \
  --deployment <deployment> \
  --input-text <input-text> \
  --resource-group <resource-group> \
  [--user <user>] \
  [--encoding-format <encoding-format>] \
  [--dimensions <dimensions>]
Parameter Type Required Description
resource-name string Yes The name of the Microsoft Foundry resource.
deployment string Yes The name of the deployment.
input-text string Yes The input text to generate embeddings for.
resource-group string Yes The Azure resource group name.
user string No User identifier for tracking and abuse monitoring.
encoding-format string No The format to return embeddings in (float or base64).
dimensions string No The number of dimensions for the embedding output. Only supported in some models.

Tool annotation hints:

Destructive Idempotent Open World Read Only Secret Local Required

OpenAI: Create completions

Create text completions by using Azure OpenAI in Foundry. Send a prompt or question to Azure OpenAI models deployed in your Foundry resource and receive generated text answers. Use this feature when you need to create completions, get AI-generated content, generate answers to questions, or produce text completions from Azure OpenAI based on any input prompt. Supports customization with temperature and max tokens.

Example CLI command

azmcp foundryextensions openai create-completion \
  --deployment <deployment> \
  --prompt-text <prompt-text> \
  --resource-name <resource-name> \
  --resource-group <resource-group> \
  [--max-tokens <max-tokens>] \
  [--temperature <temperature>]
Parameter Type Required Description
deployment string Yes The name of the deployment.
prompt-text string Yes The prompt text to send to the completion model.
resource-name string Yes The name of the Microsoft Foundry resource.
resource-group string Yes The Azure resource group name.
max-tokens string No The maximum number of tokens to generate in the completion.
temperature string No Controls randomness in the output. Lower values make it more deterministic.

Tool annotation hints:

Destructive Idempotent Open World Read Only Secret Local Required

OpenAI: List models and deployments

List Azure OpenAI model deployments in a Microsoft Foundry resource. This tool retrieves information about deployed models including model names, versions, capabilities, and deployment status.

Example CLI command

azmcp foundryextensions openai models-list \
  --resource-name <resource-name> \
  --resource-group <resource-group>
Parameter Type Required Description
resource-name string Yes The name of the Microsoft Foundry resource.
resource-group string Yes The Azure resource group name.

Tool annotation hints:

Destructive Idempotent Open World Read Only Secret Local Required

Resources: Get Foundry resource

List or get Microsoft Foundry resources and return resource-level details such as endpoint URL, location, SKU, kind, and provisioning state. If you provide a resource name, the tool returns that single resource. Otherwise, it returns the resource inventory in scope. For model deployment inventory inside a Foundry resource, use the OpenAI models list tool.

Example CLI command

azmcp foundryextensions resource get \
  [--resource-name <resource-name>]
Parameter Type Required Description
resource-name string No The name of the Microsoft Foundry resource.

Tool annotation hints:

Destructive Idempotent Open World Read Only Secret Local Required