CLI support for AWS Bedrock Integration

Ideas Feature Requests
,
1

Feature request for product/service

Cursor CLI

Describe the request

Support routing AI requests through AWS Bedrock when using the Cursor CLI. Right now the bedrock models I use in the Cursor IDE are not showing as available in the CLI after configuring the AWS Bedrock integration.

2 Likes
3

+1

I’m running the Cursor CLI (agent) in GitHub Actions for automated code analysis. My team has AWS Bedrock configured in our Cursor org (team-role mode), and it works locally. However, it does not work in headless CI.

What works locally:

  • agent login (browser auth, one-time) stores tokens in macOS Keychain
  • export CURSOR_API_KEY=
  • agent -p --model us.anthropic.claude-sonnet-4-5-20250929-v1:0 “prompt” → works

What fails in CI (GitHub Actions, Linux):

  • CURSOR_API_KEY is set via GitHub secrets
  • ~/.cursor/cli-config.json is written with full bedrock config and authInfo
  • Same model, same API key → Cannot use this model: Bedrock models are disabled for this user.

What I’ve tried:

  • Writing cli-config.json with bedrock config + authInfo block manually
  • Running agent bedrock use-team-role in CI (subcommand not available without login tokens)
  • Various combinations of --trust, running from different directories, etc.

Request: Please support Bedrock model access via API key authentication alone, without requiring a prior browser login. This would enable headless CI/CD use cases with Bedrock models, which is a core use case for the API key feature.

4

Any update on this? Is Cursor CLI feature parity a planned development? The IDE allows you to select custom models - ex via Bedrock - but the cli no. Furthermore the workaround showed does not work even tho I can use bedrock via IDE:

agent -p --model eu.anthropic.claude-opus-4-6-v1
Cannot use this model: Bedrock models are disabled for this user.