An exciting open-source project is making waves: openclaw-a365 connects the popular AI agent runtime OpenClaw directly with Microsoft 365 Agents (A365). The result: your AI agent lives natively in Microsoft Teams and Outlook – with full access to calendar, email, and user data via the Microsoft Graph API.
What is openclaw-a365?
openclaw-a365 is an open-source channel plugin (MIT license) that registers OpenClaw as a native Microsoft 365 Agent. Instead of running OpenClaw only locally on a workstation, the agent becomes a full participant in your Microsoft 365 environment.
Key Features
Native A365 Integration – The agent receives and sends messages directly through Microsoft 365 Agents. Users chat with the agent like a colleague in Teams.
Graph API Tools – Built-in tools for calendar operations (read, create, update, delete events, find meeting times), email (send), and user information (profile lookup).
Agentic Identity – The agent gets its own user identity in the tenant (e.g., agent@company.com). All actions are clearly attributed to the agent – not the user.
Multi-Model Support – Configurable LLM models with fallback support: Anthropic Claude, OpenAI GPT-4o, Azure OpenAI, or OpenRouter.
Network Policies – Control over the agent's outbound network access: unrestricted, restricted (Microsoft + LLM provider only), or allowlist (custom domains).
Role-Based Access – Distinction between Owner (primary user) and Requester (other users) with different permissions.
Agentic Identity: The Key Differentiator
A core concept of openclaw-a365 is the Agentic Identity. Unlike traditional OAuth flows where an app acts "on behalf of the user," the agent has its own identity:
Traditional Approach: The agent acts as the user and has access to everything the user can see. Audit logs show "user did X via app."
Agentic Identity: The agent acts as itself (agent@company.com) and only has access to explicitly shared resources. Audit logs clearly show "agent@company.com did X."
This means: The owner shares their calendar with the agent – just like sharing it with a colleague. The agent only sees what has been explicitly granted.
Network Policies: Container-Level Security
A particularly important feature for enterprise environments are the Network Policies. Since AI agents can execute code that makes network requests, openclaw-a365 offers three modes:
Unrestricted – No restrictions (default). Suitable for development and testing.
Restricted – Only essential domains allowed: Microsoft authentication, Graph API, and the configured LLM provider. Everything else is blocked.
Allowlist – Essential domains plus custom domains you specify. Ideal for production environments with controlled API access.
Enforcement happens via iptables at the container level and captures all outbound requests – including shell commands, curl, or any code the LLM generates.
Deployment: Docker and Azure
openclaw-a365 is delivered as a Docker container and is ideal for hosting on Microsoft Azure:
Azure Container Instances (ACI) – Fastest option for individual agents. Containers start in seconds and require no cluster management.
Azure Kubernetes Service (AKS) – For enterprises looking to scale multiple agents. Full orchestration, auto-scaling, and rolling updates.
Azure Container Apps – The middle ground: serverless containers with automatic scaling and built-in HTTPS. Ideal for teams wanting the simplicity of ACI with AKS scalability.
Azure Virtual Machines – For maximum control and isolation. Particularly recommended by Microsoft for security-critical agent deployments.
In all scenarios, you benefit from proximity to Microsoft Entra ID (authentication), Azure Key Vault (secrets management), and Azure data centers in Switzerland.
How CNEXT Supports You
As a Microsoft partner and OpenClaw specialist, CNEXT offers the complete package for your AI agent project:
Consulting & Architecture
We analyze your requirements and design the optimal agent architecture – from identity strategy and network policies to cloud infrastructure.
Implementation & Deployment
We set up openclaw-a365 in your Azure environment, configure the Microsoft 365 Agent registration, create the Agentic Identity, and implement appropriate network policies.
Security & Compliance
Building on our security expertise (see our article on OpenClaw security risks), we secure your agent with Microsoft Defender, Conditional Access, and continuous monitoring.
Custom Skills & Extensions
We develop tailored skills and Graph API integrations for your specific business processes – from CRM connections to internal data sources.
Azure Hosting & Operations
We handle hosting on Azure with Swiss data centers, including monitoring, updates, and support. You focus on usage – we take care of operations.
Conclusion
openclaw-a365 brings the flexibility of OpenClaw into the enterprise world of Microsoft 365. With Agentic Identity, Network Policies, and Azure hosting, you can build a secure, controllable AI agent that integrates seamlessly into Teams and Outlook. CNEXT guides you from strategy to live operations.
Contact us for a demo or consultation.