Anthropic introduces Claude Managed Agents to accelerate AI agent development

Anthropic PBC today launched Claude Managed Agents, a cloud service that customers can use to build artificial intelligence agents.

The company says this offering shortens the development workflow from months to weeks.

Deploying a production-level agent requires software teams to build not only the agent itself but also a significant amount of scaffolding. Developers must configure a container in which the agent can perform tasks without compromising production systems. Additionally, they have to set up the infrastructure on which the container will run, visibility features and other things.

Claude Managed Agents automate many of those tasks. It is accessible through a collection of links to organize applications. According to Anthropic, customers will be billed by the Claude model for the use of their agents and a fee of eight cents per hour of agent startup time.

Developers can start a Claude Managed Agents project by providing a description of the tasks they wish to automate with an AI agent. From there, they must specify the tools, or third-party applications, that the agent must use to perform the specified tasks. They can also define cybersecurity rules to control information such as whether a tool can be activated without the user’s permission.

Agents automatically use one container per agent. That container holds the software components predefined by the agent developer. For example, a company that develops a web design agency may wish to equip its sandbox with a browser.

Claude Managed Agents automate much of the work involved in state management. That is the data management function that AI agents use to complete tasks. A coding assistant, for example, might incorporate programming advice from the public web into its quick responses. State management also includes highly sensitive data such as login credentials where an agent logs into cloud-managed tools.

Instrument orchestration is another task that Managed Agents promises to make easy. When the agent receives information, the service determines which tools are available to use to generate a response. A fault-detection mechanism enables agents to continue where they left off after an interruption in their operation.

Two Affiliates features are currently in research preview. The first one enables an agent to coordinate with other agents when working on complex tasks. Another feature is configuring Claude to automatically refine the quality of the fast response. The company claims that the storage capability “improved task efficiency by 10 points over a standard feedback loop” in internal testing.

Claude Managed Agents’ early user base includes high-profile technology firms such as Notion Inc., Rakuten Group Inc. and Asana Inc. Anthropic says several customers have already integrated AI agents built using the service into their products.

Image: Anthropic