OpenAI, Microsoft, Google, and Anthropic All Launch Enterprise Agent Platforms in April 2026
Summary
- • OpenAI, Microsoft, Google, and Anthropic each launched or updated enterprise AI agent platforms within weeks of each other in April 2026.
- • Microsoft Foundry adds per-session sandboxes, E2B Code Interpreter, and OpenTelemetry; Google launches Gemini Enterprise Agent Platform with multi-agent orchestration.
- • OpenAI workspace agents (Codex-powered) run in the cloud and are shareable via Slack; Anthropic Claude Managed Agents entered public beta in early April.
- • Analysts say the wave marks a shift from front-end chatbots to agents that can spin up dedicated containers and scale to zero.
Details
OpenAI launches ChatGPT workspace agents powered by Codex
Described as an evolution of GPTs, workspace agents can prepare reports, write code, respond to messages, and run in the cloud continuously. They are shareable within organizations via ChatGPT or Slack. Analyst Jason Andersen noted OpenAI is late to this capability relative to Google, Microsoft, and Anthropic.
Microsoft Foundry Agent Service hosted agents with per-session sandboxes and scale-to-zero
Following a preview at Microsoft Ignite, the production release includes secure per-session sandboxes with filesystem persistence, integrated identity, scale-to-zero economics, a Code Interpreter plugin with E2B integration, security controls, and OpenTelemetry observability. Microsoft described it as 'a fundamentally different experience' from the preview.
Google launches Gemini Enterprise Agent Platform with multi-agent orchestration
Launched the same day as OpenAI and Microsoft's announcements, the platform features multi-agent orchestration, agent-to-agent communication, natural language interfaces for non-technical users, and evaluation frameworks. Google also updated the Gemini Enterprise app for office workers to build, manage, and interact with agents.
Anthropic Claude Managed Agents in public beta — composable APIs for cloud-hosted agents
Introduced in early April 2026, Claude Managed Agents is a suite of composable APIs for building and hosting cloud-hosted agents. It preceded the other three announcements and was cited as the context-setter that the OpenAI, Microsoft, and Google releases followed within weeks.
New platforms enable agents to spin up dedicated containers and scale to zero
A key architectural shift separates these platforms from earlier approaches: rather than treating agents as front-end interfaces, they provide dedicated container compute that agents can provision dynamically and release when idle, reducing cost. This infrastructure tier is designed to handle agents that write and execute code as part of task completion.
Analyst: rising agent capability increases complexity and governance needs
Jason Andersen (Moor Insights & Strategy) argued that as agents become more capable — including writing code and executing multi-step tasks — managing them well becomes more complex. He characterized the managed hosting options from Microsoft and Anthropic as the infrastructure response to that challenge, distinct from simple front-end deployments.
Product Launch = new product or feature release, Infrastructure = compute or hosting architecture shift, Insight = analyst interpretation
What This Means
Enterprise agentic infrastructure is rapidly maturing from prototype tooling into production-grade platforms with security sandboxes, observability, and multi-agent orchestration — AI practitioners evaluating agent deployment options now have competing offerings from all four major vendors to assess simultaneously. The convergence also signals that managed hosting with proper isolation and governance is becoming table stakes for production agents, not a differentiator.