AWS Shows How to Pair OpenCode With Open-Weight Models on Amazon Bedrock

On September 23, 2026, AWS published a technical guide to pairing OpenCode with open-weight models on Amazon Bedrock. It focuses on setup and workflow choices rather than announcing a new Bedrock service. AWS had described a Bedrock console experience for selecting coding agents such as OpenCode and routing requests through Bedrock in a June post; the new guide focuses on putting OpenCode and open-weight models together in practice.

AWS describes OpenCode as an open-source, terminal-native agent built in Go that can read and edit files, run shell commands, and use Language Server Protocol diagnostics to understand projects. AWS says it connects to more than 75 model providers, including Bedrock. In the setup described, OpenCode runs locally while inference takes place within the user’s AWS account. AWS also says the approach requires no infrastructure management, has no per-seat fees, and is pay-per-use.

The guide demonstrates multi-model workflows with Moonshot AI Kimi K3, OpenAI GPT-OSS 120B, and NVIDIA Nemotron 3 Super 120B, showing how models can be matched to different tasks. AWS also cites Ethara.AI as a production example, saying the company uses the architecture with multi-agent orchestration for AI engineering and research workflows. For development teams, the practical value is the guidance on connecting Bedrock model choices to a terminal-based coding workflow; the account, data-handling, and billing characteristics remain AWS’s descriptions of the approach.