Moonshot AI has enhanced its open-source coding agent, Kimi Code CLI, to integrate directly with Amazon Bedrock, enabling users to leverage Amazon Web Services (AWS) standard credential chain for authentication. This update, introduced in version 2.7.0, eliminates the need for a separate Amazon Bedrock API key, simplifying the connection process for AWS customers.
Streamlined Authentication for Developers
Information was available with The Chenab Times indicating that the integration allows Kimi Code CLI to utilize existing AWS configurations, such as named profiles and single sign-on (SSO) credentials. Previously, users had to set up a generic connection provider and manage a distinct Amazon Bedrock API key, adding an extra layer of complexity and security overhead.
The new native bedrock provider within the CLI resolves temporary AWS credentials through the standard AWS Command Line Interface (AWS CLI) credential chain. This means developers can now employ their established AWS authentication workflows, including those managed by AWS IAM Identity Center (successor to AWS SSO), without requiring separate credential management for the Kimi Code agent.
Enhanced Functionality and Model Access
This advancement facilitates access to models on Amazon Bedrock, such as Kimi K3.1 Pro Highspeed, which supports advanced features including streaming, configurable reasoning, image input, and client-side tool calling. Amazon Bedrock itself offers pay-per-token inference for a variety of AI models from leading providers, operating within the AWS data boundary.
The native provider simplifies the setup by automatically resolving the appropriate regional endpoint for the model and generating short-lived bearer tokens using the @aws/bedrock-token-generator package. This process ensures that sensitive AWS access keys or manually generated Amazon Bedrock keys are not required to be embedded within the Kimi Code configuration file.
Developers will need an AWS account with access to the target model and inference profile in a supported AWS Region, Kimi Code CLI version 2.7.0 or later, and the AWS CLI installed. Appropriate AWS Identity and Access Management (IAM) permissions are crucial, granting the identity used by Kimi Code CLI the ability to invoke the selected models and inference profiles.
Configuration and Best Practices
To configure the new provider, users are advised to directly edit the Kimi Code CLI configuration file (~/.kimi-code/config.toml) rather than relying solely on the desktop or web interfaces, as certain environment fields are not yet exposed through those UIs. Explicitly declaring model capabilities, such as reasoning, image input, and tool use, within the configuration file is also recommended to ensure Kimi Code can fully utilize the model’s features.
Consistency in Region settings is paramount, and the CLI requires an explicit AWS_REGION or AWS_DEFAULT_REGION to be set, as it cannot infer this from the AWS CLI configuration alone. For optimal performance and cost-effectiveness, using global inference profiles is encouraged where data-residency requirements permit.
The article also provides guidance on overriding endpoint URLs and protocols if custom configurations, such as those involving VPC endpoints or specific wire protocols like OpenAI Responses or Anthropic Messages, are necessary. It emphasizes treating the connectivity status reported by Kimi Code as an indicator, recommending a non-interactive prompt test to validate the full authentication and invocation path.
Simplified Workflow and Security Considerations
The integration aims to streamline the developer workflow by leveraging existing AWS security infrastructure. While the native provider reduces reliance on API keys, users retain responsibility for protecting their AWS credentials and adhering to the principle of least privilege when assigning IAM permissions. Using dedicated IAM roles for Kimi Code CLI is suggested for enhanced security, particularly in production environments.
For cleanup, users are advised to remove any newly created IAM identities or permissions if they were set up specifically for testing. Reverting local Kimi Code CLI configurations or AWS CLI profiles may also be necessary. Since Amazon Bedrock inference is charged on a per-token basis, there are no persistent cloud infrastructure deployments to manage after use.
This update by Moonshot AI signifies a move towards tighter integration between AI development tools and cloud platforms, making advanced AI models more accessible and manageable for a broader range of developers operating within the AWS ecosystem.

The Chenab Times News Desk

