AWS teaches Strands Agents to work with any model on SageMaker
AWS has published a guide to building custom parsers for the Strands Agents framework. The parsers let it integrate any LLM deployed on SageMaker AI, including models with no native support for the Bedrock Messages API. The example deploys Llama 3.1 through SGLang using the ml-container-creator tool. This gives enterprise AWS users a route to building agent systems on practically any open model, without being limited to the Bedrock ecosystem.
AI-processed from AWS Machine Learning Blog; edited by Hamidun News
The enterprise AI agent world has hit a moment comparable to the arrival of universal adapters for electronics: Amazon Web Services has shown how to make its Strands Agents framework work with literally any language model deployed on SageMaker, even if that model has no idea what the Bedrock platform's standard API is.
To understand why this step matters, it helps to go back a few months. AWS launched Strands Agents as an open framework for building AI agents: programs that can plan actions on their own, use tools and solve multi-step tasks. The framework was originally built around models available through Amazon Bedrock, which created a convenient but closed ecosystem. Companies that wanted to use their own fine-tuned models or open LLMs such as Llama deployed on SageMaker AI endpoints ran into a wall of incompatible formats.
The new guide, published on the AWS Machine Learning blog, tackles exactly this problem. Amazon's engineers describe in detail how to build so-called custom model parsers: intermediate layers that translate Strands Agents requests into a format a specific model understands, and convert the model's responses back into the structure the framework expects. In essence, it is a translator between two systems that speak different languages.
As a practical example, AWS demonstrates deploying Llama 3.1 with SGLang, a high-performance engine for language model inference, on SageMaker infrastructure. To simplify containerization, it uses the ml-container-creator tool from AWS Labs, which automates the creation of Docker containers for ML models. Once the model is deployed, the developer implements a custom parser that intercepts Strands Agents calls, reformats them from the Bedrock Messages API into a format compatible with the SGLang endpoint, receives the response and converts it back. Technically it is not rocket science, but without clear documentation and examples the process could turn into days of debugging.
Why does this really matter? Because in an enterprise setting, one model rarely solves every task. Companies train specialized models on their own data, experiment with open architectures and combine several models in a single workflow. Until now, using such models in agent scenarios on AWS required either switching to Bedrock-compatible formats or writing your own orchestration from scratch. Now there is a standardized way to connect practically any LLM to the agent framework while keeping all the advantages of Strands: tool management, reasoning chains and error handling.
The move fits into AWS's broader strategy of turning its ML platform into as open an ecosystem as possible. Amazon has clearly recognized that trying to lock customers into Bedrock is counterproductive in a world where new open models appear every week and corporations are investing more and more in their own fine-tuned solutions. Rather than compete with every new model, AWS offers infrastructure that can run anything and, now, an agent framework that knows how to work with that "anything."
The competitive context is worth noting too. Google, with Vertex AI Agent Builder, and Microsoft, with AutoGen and Semantic Kernel, are actively developing their own agent platforms. But AWS's approach of open custom parsers looks more flexible: instead of dictating which models to use, Amazon provides tools for integrating any of them. That may become a deciding argument for enterprise customers who do not want to depend on a single model provider.
Looking ahead, we can expect ready-made parsers for popular open models and inference engines to appear; the community will almost certainly start building a library of compatible configurations. For the AI agent market, this is another step toward maturity: the technology is no longer a privilege of those who use a specific model from a specific provider, and becomes available to any team ready to deploy an LLM on cloud infrastructure.
- Nov 19, 2026Two pods share one A10G GPU on Amazon EKS: NVIDIA Run:ai gave each 8 GiB
- Nov 19, 2026Anthropic goes public and raises $81.3 billion, beating SpaceX's $75 billion record
- Nov 19, 2026Amazon and Google's Anthropic Stakes Valued at $308 Billion in IPO at $83 a Share
- Nov 19, 2026AWS AgentCore Agents Can Now Reach a Private Keycloak Behind a Corporate CA
Test Frontier LLMs & Autonomous Agents in One Workspace
Run Claude Opus 5.5, GPT-6, Gemini 3.1, and DeepSeek side-by-side in Nomad AI (nomadnet.ai) — unified model routing, deep research, and multimodal workflows under one subscription.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.