What can Hubi see with my Langfuse keys?

Whatever the project keys allow - typically traces, observations, sessions, scores, and prompt data for that project. It reads those objects and can write scores, and it does not touch your billing or organization settings.

Can Hubi create or change evaluation scores, or only read them?

It can do both. Hubi reads existing scores and can attach new ones to a trace or observation, which is useful when you want to record a human judgment after reviewing an output.

Will Hubi pull traces from the wrong project if I have several?

No. Each set of Langfuse keys is scoped to one project, so Hubi only sees the project whose keys you connected. To work across projects you would connect each one separately.

How current are the cost and latency numbers Hubi reports?

Langfuse records traces as your application sends them, so the numbers reflect whatever has already been ingested. Very recent runs appear once Langfuse finishes processing them, which is usually quick but not always instant.

All Integrations

Langfuse

Artificial Intelligence Developer Tools

Langfuse captures every LLM trace, but spotting which prompt version regressed usually means scrolling through the dashboard. Hubi searches traces, pulls cost and latency for a given run, reads scores on a session, and reports which generations failed - so you can ask about your app's behavior instead of clicking through it.

18 tools available

USE CASES

How businesses use Langfuse + Hubi AI

Real automation scenarios powered by this integration.

🔍

Find the trace behind a bad answer

Give Hubi a session or user id and it pulls the matching Langfuse traces, observations, and the prompt and response for each step.

Get TraceList ObservationsGet Session

💰

Check cost and latency for a run

Ask what last night's batch cost and Hubi reads token usage, model, and latency from the traces and sums it into a plain answer.

List TracesGet TraceGet Observation

📝

Score a trace from a review

After you judge an output, Hubi attaches a Langfuse score to that trace or observation so the evaluation lives next to the data.

Create ScoreGet TraceList Scores

🚨

Flag failing generations in Slack

When traces show errors or low scores, Hubi gathers them and posts a summary to your team channel so regressions surface fast.

List TracesList ScoresSend Message

HOW IT WORKS

Get started in 3 steps

No workflows to build. Just connect, chat, and automate.

Connect Langfuse

Authorize your account with one click. Hubi AI gets secure access to 18 tools.

Talk to your agent

Describe what you need in plain language. Your AI agent understands context and picks the right tools.

Watch it work

The agent reads, sends, labels, and organizes on your behalf - no manual setup required.

Everything you need to know about Langfuse + Hubi AI

The Hubi AI Langfuse integration connects your tools and data to your AI marketing agents. Start automating workflows in minutes.

Both, in principle. Hubi authenticates with the public and secret API keys from a Langfuse project, and those keys work the same whether the project lives on Langfuse Cloud or your own self-hosted deployment, as long as the host is reachable.

Ready to connect Langfuse to Hubi AI?

Start automating your workflows in minutes. No credit card required.