Skip to main content
Kubit works with Langfuse in two ways. If your app sends spans to Langfuse, the agent plugin adds KubitSpanProcessor next to the Langfuse processor and Langfuse keeps receiving every span. If you self-host Langfuse, Kubit can read its traces in place from your ClickHouse. This page helps you pick one and points to the steps.

Pick a Path

Next to Langfuse in Your App

The /kubit-integrate skill in the agent plugin detects Langfuse as the place your traces go today. It adds KubitSpanProcessor next to Langfuse’s processor, so Langfuse keeps receiving every span and Kubit gets a copy. Run it in Claude Code or Cursor. Agent plugin has the steps. The extraction mapping template default-agent reads Langfuse keys such as langfuse.user.id, langfuse.session.id, langfuse.release and langfuse.environment. A workspace created by /kubit-integrate starts on default-otel, which reads only the OpenTelemetry GenAI keys. Check the extraction mapping shows how to switch.

Self-Hosted Langfuse on ClickHouse

A self-hosted Langfuse stores its traces in your ClickHouse, so Kubit can query them there.
  1. Connect your ClickHouse. See Connect ClickHouse.
  2. Create a workspace with the Langfuse origin (Langfuse traces in ClickHouse) and pick the ClickHouse connection.
    Create Workspace dialog with Workspace Name Support Agent, Workspace Timezone Europe/Athens (UTC+2*) and Origin Langfuse, with the hint Langfuse traces in ClickHouse, above the Continue button.
  3. The workspace opens on the Langfuse preset tab in the Modeler. Fill in the fields below, review Tables to create, then click Validate and use Langfuse.
Under Tables to create, check Price table and Evaluator result tables. The Evaluator result tables let the Evaluator run and save its labels. Created in shows where they go, and you can pick another location.
The Schema tab stays grayed out until you apply the preset, with the tooltip Apply the Langfuse preset first. Mapping a table by hand would turn it off for good. Once a table is mapped by hand, the preset is off for that workspace and can’t be turned back on.
Langfuse Cloud keeps traces in Langfuse’s own storage, which Kubit can’t read in place. Use the agent plugin path instead.

Quick Test With a Langfuse Export

To try Kubit on traces you already have, export Langfuse observations to a JSONL file and upload it to a Sandbox workspace. Kubit reads no scores from the observations, so add them as a separate labels file. Upload traces to a Sandbox workspace has the export commands and the upload steps.

Next Steps

  • Agent plugin: the steps for adding Kubit next to Langfuse.
  • Modeler: presets and table mapping for warehouse workspaces.