Announcing Azure Language in Foundry Tools for Deterministic, Privacy-First Agents
Xiaoying breaks down the new Azure Language features in Foundry Tools, highlighting privacy-first AI agents, advanced PII redaction, and deterministic intent routing for developers building enterprise solutions on Azure.
Announcing Azure Language in Foundry Tools for Deterministic, Privacy-First Agents
As the demand for robust and compliant AI agent architectures accelerates, Microsoft introduces new Azure Language capabilities within Foundry Tools. With a focus on privacy, security, and determinism, these tools equip developers to build agents that meet enterprise requirements for predictability and compliance.
Key Enhancements
Azure Language Remote MCP Server
- Centralized Language AI Tools: Access a suite of API-driven Language AI services (PII redaction, intent detection, entity extraction, question answering, and more) through a cloud-hosted Model Context Protocol (MCP) server, simplifying agent workflows.
- Seamless Integration: Native support for Foundry resource endpoints, two authentication methods (key-based and Microsoft Entra ID/RBAC), and straightforward addition via Foundry portal or agent playground.
- No infrastructure management: Cloud-hosted, fully managed, and agent-ready.
Supported Language AI Capabilities
- PII Redaction: Protect privacy in text or native documents by detecting and masking personal identifiers.
- Intent Detection (CLU): Accurately route user queries by extracting intents and entities from conversational messages.
- Exact Question Answering: Retrieve grounded answers from configured knowledge bases, supporting compliance-heavy scenarios.
- Named Entity Recognition & Healthcare Entity Extraction: Identify people, organizations, medical entities, and financial terms.
- Sentiment Analysis & Opinion Mining: Monitor sentiment levels for customer support or stakeholder feedback workflows.
- Text Summarization & Key Phrase Extraction: Highlight actionable insights from meetings, documents, and conversations.
- Language Detection: Support multilingual workflows and onboarding for users in various locales.
Example Agent Patterns
- Privacy-Preserving AI Assistants: Use PII Redaction for healthcare, claims, or sensitive HR workflows.
- Case Routing: Leverage Intent Detection and sentiment analysis to automate customer inquiries and escalation.
- Enterprise Knowledge Assistants: Deliver reliable, policy-grounded responses using knowledge-based QA.
- Compliance & Operations: Turn large documents into structured insights with NER, summarization, and key phrase extraction.
- Multilingual HR Assistants: Detect languages and extract relevant entities for global teams.
- Meeting/Productivity Assistants: Summarize key decisions and follow-ups from transcripts or documents.
Getting Started
- Connect to MCP Endpoint
- Example:
https://{foundry-resource-name}.cognitiveservices.azure.com/language/mcp?api-version=2025-11-15-preview
- Example:
- Authenticate
- Use key from Foundry resource (for prototyping).
- Use Microsoft Entra ID for RBAC in enterprise scenarios.
- Configure in Foundry Portal
- Add from Foundry Tools catalog, or directly to agents via agent playground.
Integration in VS Code with GitHub Copilot
- Connect Azure Language MCP server in VS Code (using mcp.json and correct headers).
- Test redaction, intent, and other AI tools directly from your developer environment.
Advanced Privacy & PII Redaction
- Improved Model (2025-11-15-preview): Supports more entity types (Airport, City, Geopolitical Entity, etc.) and enhanced accuracy.
- Synthetic Replacement: Mask PII with realistic synthetic values for safer sharing and analysis.
- Analytics: Confidence scores, entity offset, and detailed tags for granular review.
- Customer Validation: Real-world improvements with users like Nationwide Building Society (over 90% redaction accuracy).
Deterministic Intent Routing with CLU
- Multi-turn Understanding: Accurately interpret full conversation context for better intent prediction.
- Slot-Filling: Map entities to intents, identify missing information, and enable smarter bot interactions for processes like flight booking.
- Quick Deploy with LLM: Streamlined deployment and playground testing for custom CLU models.
Summary
With Azure Language in Foundry Tools, developers gain access to a scalable, privacy-first Language AI foundation built for modern agentic solutions. Whether you’re automating customer service, compliance tasks, or knowledge workflows, these tools enable confident innovation on Azure.
Learn More:
- Azure Language in Foundry Tools
- Quickstart: Multi-turn conversational language understanding with CLU
- Foundry Playground
- What’s New in Azure Language
This post appeared first on “Microsoft AI Foundry Blog”. Read the entire article here