Microsoft Copilot agents represent the company’s answer to the agentic AI era, and now Microsoft is taking direct control of the technology powering them. The company is developing its own secure, proprietary version of OpenClaw—an open-source AI agent framework—specifically for integration with Copilot to address security vulnerabilities and unlock more proactive automation capabilities across Microsoft 365.
Key Takeaways
- Microsoft is building a proprietary OpenClaw variant led by a team under an individual named Shahine to power more secure Microsoft Copilot agents.
- OpenClaw, created by Peter Steinberger (now at OpenAI), runs locally to manage files, send emails, browse the web, and automate workflows.
- Microsoft 365 Copilot already supports custom agents, Copilot Search, and integration with Anthropic’s Claude Cowork for multi-step workflows.
- OpenClaw has the most traction among at least four competing AI agent frameworks despite unresolved security concerns.
- A preview of Microsoft’s OpenClaw-inspired features may debut at Build 2026 in June in San Francisco.
Why Microsoft is building its own Microsoft Copilot agents framework
The original OpenClaw framework, despite its popularity, carries security risks that have prompted Microsoft to invest in a proprietary alternative. OpenClaw has gained the most traction among competing AI agent frameworks, but security concerns shadow its adoption. By building its own version, Microsoft can integrate tighter controls, ensure compliance with enterprise standards, and make Microsoft Copilot agents more reliable for business-critical workflows.
Microsoft 365 Copilot already supports custom agents that can automate tasks, analyze data, make decisions, and adapt over time based on interactions. However, the current implementation relies partly on third-party frameworks. Building an in-house secure variant of OpenClaw would let Microsoft control the entire stack—from data handling to task execution—without exposing sensitive business information to external infrastructure.
The architectural difference between OpenClaw and Microsoft’s cloud-based Copilot approach is significant. OpenClaw runs locally as a self-hosted solution, giving users direct control over their data and workflows. By contrast, Copilot operates as a cloud-based service that feeds user data to Microsoft servers. Microsoft’s proprietary OpenClaw variant could bridge this gap—offering the proactive automation of OpenClaw with the enterprise security and integration depth of Copilot.
What Microsoft Copilot agents can already do
Microsoft 365 Copilot has already begun supporting multiple agent types and integrations that hint at where the company is heading. The platform currently offers custom agents that users can build for specific workflows, Copilot Search for AI-powered discovery across Microsoft and third-party applications, and a Researcher agent that uses multiple AI models to improve accuracy. Additionally, Microsoft has integrated Anthropic’s Claude Cowork technology, which enables long-running, multi-step workflows that span across apps and files.
OpenClaw itself already has a fully integrated plugin for Microsoft Teams, allowing users to interact with agents via direct messages, group chats, or channels. This existing integration shows how smoothly agentic workflows can fit into familiar collaboration tools. Microsoft’s goal with its proprietary version is to expand these capabilities while removing the security friction that has kept some enterprises from adopting OpenClaw more broadly.
Timeline and competitive context
Microsoft has not announced a firm launch date for its secure OpenClaw variant, but the earliest preview may occur at Build 2026 in June in San Francisco. This timeline gives Microsoft’s team time to address the security gaps that plague the original framework while integrating the technology deeply with Microsoft 365 services.
The competitive landscape includes at least four other AI agent frameworks, but OpenClaw remains the leader in adoption. Microsoft’s move signals that the company sees agentic AI as central to Copilot’s future, especially as enterprise adoption of Copilot itself has faced monetization challenges. By owning the agent framework, Microsoft can differentiate Copilot from competitors like Anthropic’s Claude models and ensure that Microsoft Copilot agents become the default choice for Microsoft 365 customers.
How to build Microsoft Copilot agents today
While Microsoft’s proprietary OpenClaw variant is still in development, businesses can already start building custom agents within Microsoft 365 Copilot. The process begins by identifying use cases—outlining specific goals such as optimizing customer service or gaining data insights—and analyzing existing workflows for automation opportunities. Once you identify a workflow to automate, you can configure agents to handle specific tasks within your Microsoft ecosystem.
For those willing to experiment with OpenClaw directly, the setup process is straightforward. You can connect agents to tools like Microsoft To-Do through a few configuration messages, then train the agent on desired behaviors. The ease of integration demonstrates why OpenClaw has attracted so much attention despite its security limitations.
Is Microsoft Copilot agents the future of work automation?
Microsoft certainly believes so. The company’s investment in building a secure, proprietary agent framework suggests that Copilot agents will become increasingly central to how Microsoft 365 customers automate work. Rather than relying on external frameworks or competing platforms, Microsoft is taking control of the technology stack to ensure that Microsoft Copilot agents integrate smoothly with Teams, Outlook, Word, Excel, and other core productivity tools.
The shift toward agentic AI represents a fundamental change in how software works. Instead of users manually triggering actions, agents can proactively handle multi-step workflows, learn from interactions, and adapt to changing business needs. Microsoft‘s bet is that enterprises will pay for this capability—and that owning the underlying framework gives Copilot an insurmountable advantage over competitors.
When will Microsoft’s secure OpenClaw variant launch?
Microsoft has not announced a specific launch date, but the earliest preview may occur at Build 2026 in June in San Francisco. Full availability for enterprise customers likely depends on completing security audits and integration testing with Microsoft 365 services.
How does OpenClaw compare to Microsoft Copilot agents?
OpenClaw runs locally on your infrastructure, giving you direct control over data and workflows, while Copilot operates as a cloud-based service that sends user data to Microsoft servers. OpenClaw is more private by design, but Copilot offers deeper integration with Microsoft 365 and enterprise compliance features. Microsoft’s proprietary OpenClaw variant aims to combine the best of both approaches.
Can I use OpenClaw with Microsoft Teams today?
Yes. OpenClaw has a fully integrated plugin for Microsoft Teams that allows interaction via direct messages, group chats, or channels. This means you can start experimenting with agent-based workflows in Teams right now, though Microsoft’s more secure variant will eventually replace or supplement this option.
Microsoft’s push into agentic AI reflects a broader industry shift toward autonomous systems that handle complex, multi-step work without constant human intervention. By building its own secure version of OpenClaw, Microsoft is betting that enterprises will trust Copilot more than open-source alternatives—and that the combination of proactive agents and deep Microsoft 365 integration will justify the investment.
Edited by the All Things Geek team.
Source: TechRadar
