Anthropic on Wednesday launched Claude Managed Agents, a suite of composable APIs for building and deploying cloud-hosted AI agents at scale. The service, now in public beta on the Claude Platform, handles sandboxed code execution, authentication, checkpointing, credential management, and orchestration so enterprise teams can focus on agent logic rather than infrastructure.

The pitch is speed. According to Anthropic’s blog post, Managed Agents is designed to compress the path from prototype to production “from months to days.” Teams define tasks, tools, and guardrails. Anthropic’s infrastructure decides when to call tools, how to manage context, and how to recover from errors.

What the Service Includes

Managed Agents ships with several features aimed at production workloads: secure sandboxing, scoped permissions, identity management, long-running sessions that persist through disconnections, and execution tracing built into the Claude Console. Multi-agent coordination, where agents can spin up and direct other agents in parallel, is available in research preview.

In internal testing on structured file generation tasks, Managed Agents improved success rates by up to 10 points over standard prompting, according to Anthropic, with the largest gains on the hardest problems.

Early Adopters

Five enterprise customers are already building on the platform, per Anthropic’s announcement:

  • Notion is using Managed Agents to let teams delegate work to Claude directly inside their workspace, with dozens of tasks running in parallel. The feature is in private alpha inside Notion Custom Agents.
  • Rakuten shipped enterprise agents across product, sales, marketing, finance, and HR that connect to Slack and Teams. Each specialist agent was deployed within a week, according to Anthropic.
  • Asana built “AI Teammates,” collaborative agents that work alongside humans inside Asana projects.
  • Sentry paired its debugging agent Seer with a Claude-powered agent that writes patches and opens PRs, going from flagged bug to reviewable fix in one flow.
  • Vibecode uses Managed Agents as its default integration for prompt-to-deployed-app workflows.

The Competitive Positioning

This launch shifts Anthropic from pure model provider to agent infrastructure vendor. As Startup Fortune noted, the service puts Anthropic in direct competition with AWS Bedrock Agents, Azure AI, and Google Vertex Agent Builder for the managed agent runtime layer.

The timing aligns with a broader industry convergence. Salesforce launched AgentForce. Microsoft shipped Agent Framework v1.0 last week. OpenAI has been expanding function calling and tool use. Every major player is moving from raw model access toward platforms that wrap models in production-ready tooling.

For enterprises already on Claude’s API, Managed Agents introduces a decision: continue managing your own agent infrastructure, or hand the runtime to the same vendor that provides your model. That consolidation reduces operational complexity. It also deepens vendor dependency. As The New Stack reported, once an organization’s agents, monitoring dashboards, and operational configurations run on Anthropic’s managed infrastructure, switching costs increase.

Pricing and Access

Managed Agents is available now in public beta. Anthropic has not yet published detailed pricing. For enterprises evaluating their agent stack, the pricing structure will determine whether Managed Agents pulls teams away from self-hosted solutions or remains limited to organizations with budget flexibility and low tolerance for operational risk.