AWS open-sources Chaplin: self-service AWS Health analytics built on Amazon Bedrock agents

AWS has released Chaplin, an open source project for self-service analysis of AWS Health events using AI agents built on Amazon Bedrock and the Model Context Protocol.

What is Chaplin

Chaplin stands for Customer Health and Planned Lifecycle Intelligence Nexus. It is not another paid AWS service but an open toolkit that teams can deploy themselves, adapt to their stack and extend for specific tasks.

At its core are AI agents integrated with the AWS Health API through MCP (Model Context Protocol). The agents can parse structured AWS events, such as notifications about outages, planned maintenance and resource lifecycle changes, and turn them into concrete natural-language conclusions. A summary of active incidents is now available not through the AWS Console but through a simple chat interface.

How the MCP architecture works

MCP is an open standard for communication between LLM applications and external tools. Anthropic promotes it, and major players, from Microsoft to Google, are actively adopting it. In Chaplin, Amazon Bedrock agents connect to AWS Health tools through an MCP server, which provides a flexible extension point without rewriting the core.

With this architecture, the agents can:

• Query the status of specific resources: EC2, RDS, Lambda and other services
• Filter events by region, account or service type
• Build a timeline of incidents and planned technical changes
• Produce summaries for non-technical stakeholders, without jargon
• Recommend next steps based on the event's type and priority

Any team member, whether a product manager, an SRE or a financial analyst, can ask a question in natural language and get a structured answer without learning the console or writing SQL.

Why teams need it

AWS Health generates a constant stream of notifications: planned maintenance windows, deprecation warnings, regional incidents and limit changes. Sorting through this stream used to mean working manually through dashboards or specialized tools that require separate skills. Chaplin turns this process into self-service.

Teams can now:

• Get a summary of all active events across their infrastructure in a few seconds
• Find out which resources an upcoming planned maintenance will affect
• Assess an incident's impact on a specific workload or business process
• Automatically notify the right people through the channel of their choice

Because Chaplin is published as open source, organizations can adapt it: add their own data sources, change the agent logic or integrate it with internal monitoring and ticketing systems.

Amazon Bedrock as the foundation

For the agents, AWS uses Amazon Bedrock, its own platform for working with LLMs from leading providers: Anthropic, Meta*, Mistral, AI21 and others. This allows teams to choose a model based on their requirements for cost, speed and answer quality.

The combination of Bedrock and MCP is an emerging pattern in the AWS ecosystem for enterprise AI: not just a chatbot, but an agent with real access to live infrastructure data, one that can not only answer questions but also initiate requests to external systems.

What this means

Chaplin is a telling example of MCP ceasing to be an academic standard and becoming a real production architecture. AWS uses it not in a demo but in a solution that handles critical infrastructure status data. Companies that want to cut their response time to infrastructure events and reduce the load on DevOps teams should take a look at this tool.

*Meta has been designated an extremist organization and is banned in Russia.