Both OpenAI and Anthropic have rolled out multi-agent capabilities designed to let AI systems coordinate, delegate, and execute complex workflows autonomously, with a laser focus on enterprise customers.

Anthropic launched Claude Managed Agents in public beta on April 8, 2026, offering hosted infrastructure that supports long-running autonomous agents with built-in sandboxing and multi-agent orchestration. OpenAI has enhanced its Codex and Frontier offerings with comparable multi-agent features.

From single agents to agent teams

In February 2026, Anthropic introduced “agent teams” in its Opus 4.6 model, enabling parallel task execution where multiple agents tackle different parts of a workflow simultaneously.

Anthropic reported a 90.2% performance improvement from multi-agent systems over single-agent setups during internal evaluations. OpenAI’s approach through Codex and Frontier builds infrastructure that allows multiple AI agents to work together on tasks that previously required significant human coordination.

Enterprise adoption is already underway

Notion, Rakuten, and Asana are among the companies leveraging AI agents for operational tasks spanning project management, HR, and software development. Anthropic’s Managed Agents platform prioritizes security, sandboxing, and scalability.

What this means for investors

One notable absence in all of this: neither company has referenced crypto tokens or digital assets in connection with their multi-agent launches.

Scaling multi-agent systems, where multiple AI models run simultaneously on complex tasks, requires significantly more compute than single-agent interactions. The 90.2% performance gain Anthropic claims is compelling, but the compute cost to achieve it will ultimately determine how fast adoption moves beyond early adopters into the broader enterprise market.