# Kimi Code CLI adds native Amazon Bedrock provider

Published 2026-11-20 · AI.info Team · AI.info

> Kimi Code CLI v2.7.0 can use standard AWS credentials to connect its coding agent to Amazon Bedrock models. The new provider generates short-lived bearer tokens instead of requiring users to manage a separate Bedrock API key.

Kimi Code CLI users can now connect the coding agent to [Amazon](https://aws.amazon.com/blogs/machine-learning/use-kimi-code-cli-with-models-on-amazon-bedrock/) Bedrock models using standard AWS credentials instead of managing a separate Bedrock API key. AWS described the change in a November 19 post detailing the native Bedrock provider in [Moonshot AI](https://ai.info/companies/moonshot-ai)’s Kimi Code CLI v2.7.0.

The agent is open source and was previously able to reach Bedrock models through a generic provider configured with a Bedrock API key. The new provider can use existing AWS authentication workflows, including named profiles and single sign-on.

## How authentication changes

The provider uses the AWS credential chain by default unless a Bedrock API key is configured. It uses a token-generator package to create a short-term Bedrock API key and renews it when necessary, so users do not need to put an AWS access key or manually generated Bedrock key in the Kimi Code configuration file.

Generated keys remain valid for up to 12 hours or the length of the AWS session, whichever is shorter. A Bedrock API key explicitly supplied through the configuration takes precedence over the credential-chain path.

The provider also infers an API format and regional endpoint from the model ID, although users can override them. For Kimi models, the inferred path uses the OpenAI-compatible Chat Completions API.

AWS recommends short-term credentials over long-term ones and says users remain responsible for protecting their AWS credentials and setting appropriate permissions. Requests still require successful authentication, authorization, model access and network connectivity.

## Model and cost details

AWS uses Kimi K3.1 Pro Highspeed to demonstrate the connection. The model supports streaming, configurable reasoning, image input and client-side tool calling; the CLI can stream responses and use tool calling during coding tasks.

Its example uses the global inference profile *global.moonshotai.kimi-k3.1*, while the AWS Region determines which regional endpoint receives a request. The example configuration declares a 1-million-token context window and explicitly lists reasoning, image-input and tool-use capabilities.

Bedrock charges for inference on a pay-per-token basis. The walkthrough uses global cross-Region inference for the example model, and requires an account with access to the model and inference profile in a supported source Region.

## Setup requirements

The walkthrough requires Kimi Code CLI v2.7.0 or later and an AWS identity with permission to invoke the target model and inference profile. The AWS CLI is needed for the profile and identity-check commands in the post; its optional console-credential login workflow requires AWS CLI v2.32.0 or later.

The example policy grants permissions for the specified Kimi model and global inference profile. A Region set in an AWS CLI profile applies to AWS CLI commands, but the Region must still be set explicitly for Kimi Code.

Kimi Code also does not automatically detect the capabilities of most Bedrock models, including Kimi models; the example declares them in that file.

## Sources

1. Original source · Blog post [Amazon Web Services](https://aws.amazon.com/blogs/machine-learning/use-kimi-code-cli-with-models-on-amazon-bedrock/) aws.amazon.com

## Companies in this story

- [Amazon](https://ai.info/companies/amazon)
- [Amazon Web Services](https://ai.info/companies/amazon-web-services)

## Related stories

- [Moonshot Launches Kimi K3.1 Pro With 1M-Token Context](https://ai.info/news/moonshot-kimi-k3-1-pro-release)
- [Amazon Bedrock adds search across spreadsheet tables](https://ai.info/news/amazon-bedrock-managed-knowledge-base-tabular-search)
- [OpenAI Disrupts Moonshot-Linked Campaign to Extract Model Reasoning](https://ai.info/news/openai-disrupts-moonshot-linked-model-distillation-campaign)
- [Anthropic Report Alleges Moonshot and DeepSeek Routed User Requests to Claude](https://ai.info/news/china-probes-deepseek-moonshot-anthropic-data-routing)
- [AWS details open-source stack for Bedrock WhatsApp agents](https://ai.info/news/aws-open-source-bedrock-whatsapp-agent-stack)

## Explore

[More articles](https://ai.info/news)

---

Canonical: https://ai.info/news/kimi-code-cli-native-amazon-bedrock-provider
