n8n + Lang fuse integration

[dreddy08]

Describe your question

We are running a self-hosted n8n instance without the enterprise version and are trying to implement Langfuse for tracing, observability, token usage, and related metrics. I installed the community node @langfuse/n8n-nodes-langfuse, but it currently only supports fetching prompts and does not provide other functionality. Our workflow uses an AI Agent with the AWS Bedrock Chat Model. What would be the recommended approach to properly implement Langfuse in this setup?

Langfuse Cloud or Self-Hosted?

Self-Hosted

If Self-Hosted

No response

If Langfuse Cloud

No response

SDK and integration versions

No response

Pre-Submission Checklist

• I have checked for existing issues/discussions and consulted Langfuse AI.

[inkeep-ai-assistant[bot]]

The @langfuse/n8n-nodes-langfuse community node is limited to prompt fetching and does not provide full tracing or observability capabilities for n8n workflows.(1)

For tracing AI Agent workflows with AWS Bedrock in n8n, there's an open issue regarding output tokens not being captured with ChatBedrockConverse via n8n AI Agent. The Bedrock/ChatBedrockConverse integration doesn't provide usage metadata (like output tokens and cost) in a way Langfuse expects, so those fields may appear as "undefined" or zero in your traces.(1)

Recommended approach:

If you have control over the integration code, you can manually update generations with usage details using the Langfuse SDK. Here's the pattern for Python:(1)

from langfuse import get_client

def converse_with_bedrock(...):
    response = client.converse(...)
    langfuse = get_client()
    langfuse.update_current_generation(
        output=response['output']['message']['content'][0]['text'],
        usage_details={
            "input_tokens": response['usage']['inputTokens'],
            "output_tokens": response['usage']['outputTokens'],
            "total_tokens": response['usage']['totalTokens']
        }
    )
    return response

(1)

For AWS Bedrock credentials in self-hosted setups:

You can use the AWS SDK default credential provider chain instead of static IAM credentials. This allows you to use IAM roles for EC2 instances or ECS task roles, AWS profiles, or environment variables (AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY). You can leave the AWS Access Key ID and Secret Access Key fields blank in Langfuse, and the platform will use role credentials from the environment.(2)

Current limitations:

Langfuse can only display output tokens and cost if the underlying LLM provider returns this info and the integration passes it through. If the Bedrock model or wrapper doesn't emit this metadata, Langfuse cannot show it—this is a limitation of the upstream provider or integration.(1)

Check out these potentially useful sources:

• Instrumentation
• Observability & Application Tracing
• Best practices for managing AWS Bedrock credentials

📚 Sources:

1. bug: Output tokens not captured with ChatBedrockConverse via n8n AI Agent
2. Best practices for managing AWS Bedrock credentials across multiple Langfuse projects (self-hosted setup)

---

Have another question? Just tag @inkeep.

[Lotte-Verheyden]

Hi @dreddy08,
Currently, there is no native n8n tracing integration. Please upvote the related GitHub discussion here if you are interested in this.
One workaround we've seen is using an OpenRouter node + OpenRouter broadcast, via which you can send traces to Langfuse.