# Langfuse v4.77.0 adds Amazon Bedrock to its AI Gateway and IAM role auth for S3

published: 2026-11-09T13:05:30.338Z
topic: infrastructure
source: https://github.com/langfuse/langfuse/releases/tag/v4.77.0
url: https://aiagentnews.io/stories/langfuse-v4-77-0-adds-amazon-bedrock-to-its-ai-gateway-and-iam-role-auth-for-s3

Langfuse released v4.77.0, adding an Amazon Bedrock upstream with API key auth to its AI Gateway and IAM role authentication for Amazon S3 buckets. The changelog also covers custom base URLs and extra headers for OpenAI and Anthropic connections, along with a set of fixes.

Langfuse has released v4.77.0, which adds to its AI Gateway. The release includes an Amazon Bedrock upstream that authenticates with an API key, and it lets OpenAI and Anthropic connections use custom base URLs and extra headers. Gateway web clients now connect before reporting ready, and model routes appear as badges with an overflow view.

On the storage side, Amazon S3 buckets used for blob exports and external media storage can now authenticate with IAM roles. External media storage is now an enterprise feature. Topics also gained cost tracking for executions and runs per-trace summaries on their own queue.

The changelog lists several fixes as well. Custom base URLs are now blocked from session replay, and the storage trust policy can be copied before the external ID is displayed. Evaluators also stop when they receive a bare HTTP 403, matching how Topics already handles it.

## Why it matters if you build agents

Teams routing model calls through the AI Gateway can now add Bedrock with an API key and point OpenAI or Anthropic connections at custom endpoints with extra headers. Teams exporting to S3 or using external media storage can use an IAM role instead of long-lived keys. Check any gateway setup that relies on session replay, since custom base URLs are now excluded from it.