AWS has added agent skills for its messaging services to the AWS MCP Server, and the announcement is as short as an AWS release note gets. In AWS End User Messaging and Amazon SES now offer AI agent skills for the AWS MCP Server, dated September 25, the pitch is that developers can “build and send messages by asking their AI coding agent in plain language.” Each skill “provides your agent with step-by-step, validated guidance for tasks like verifying sending identities, building branded Rich Communication Services (RCS) agents, and sending production emails.” The example is the whole onboarding path: “you can ask Amazon SES to verify a sending identity and send your first production email.”
There are no figures, no regions, and no pricing, which is normal for a what’s-new post. What it does say is where the skills run and how they arrive. They integrate with Claude Code, Codex, Cursor, and Kiro, and in the first three “the aws-core plugin bundles the server and a curated set of skills in one install,” with other agents adding the server by configuration. The before state is stated plainly: “Previously, developers had to navigate multiple documentation pages and management console screens to complete messaging tasks.” The skills are the documentation, restructured as procedure, delivered through the same MCP server that exposes the API. That is the pattern this series has watched arrive provider by provider, and AWS is applying it to the service most likely to be driven by an agent on a schedule.
The catalog agrees with the release, which is worth saying when it happens. The Amazon SES provider page lists 13 API pages, and the verify-and-send path in the announcement runs through the Amazon SES Email API. On the Agent Readiness score both dimensions the release is about are lit: the MCP server and agent skills, alongside error semantics, OpenAPI examples, the rate limit signal, and a described event surface. The agentic access profile maps 86 operations, 55 of them acting, which is a lot of ways for an agent to send something.
The Kin Score is 80.5, exemplar band, carried by access clarity at 100.0, developer ergonomics at 86.9, and operational transparency at 84.2, with contract quality at 70.1 and contract governance at 45.5. The Agent Readiness score is 38.1, agent-ready, and what holds it there is the pair this series keeps finding on messaging providers: idempotency is unlit, and so is reversibility. An email is the canonical irreversible call, and 55 acting operations with no described idempotency guarantee is the gap a skill that says “send your first production email” should be most careful about. The identity layer is dark too, delegated identity, protected resource metadata, and dynamic client registration, so the agent still arrives with a pasted credential. AWS has given the agent the procedure. The contract does not yet say what happens when the agent runs it twice.