Prerequisite: Third-party connectors are available only on the Sana Enterprise tier, with one exception: the Workday connector is included for both Sana Core and Sana Enterprise.
Introduction
Connect Langfuse to Sana to inspect how your LLM application actually behaved — walk the observations in a trace, check evaluation scores, and leave comments on what needs a second look.
In summary, the connector has the following key characteristics:
Categories: Engineering, Analytics
Connector type: Real-time
Auth type: API keys
Hosting type: Managed
Tools
Tools are the operations the agent can use when you ask Sana about Langfuse. In practice, the agent may call more than one to complete a single request.
Tool |
Add feedback |
Get trace by ID |
List observations |
List project ID options |
List scores |
List sessions |
Log trace |
Note: This article lists the connector's full set of tools. Your workspace owner can disable individual tools for your workspace, so some of those listed here may not be available to your agents. If a tool you expected is missing, reach out to your workspace owner.
Triggers
While tools are the actions an agent takes on your behalf, triggers are the events that can start an agent automatically. When you build an agent, you can choose one of the triggers below as its starting point. For example, having the agent run every time something new happens in Langfuse.
Trigger |
New Score Received |
New Trace Received |
Scope and permissions
This connector uses API keys. When you connect your account, we securely store your keys to connect to this connector's APIs. The scopes for this connector depends on the scopes you assign your API key in Langfuse when generating the key.
Set up instructions
Prerequisites:
You are a Sana Enterprise user – all third-party connectors except the Workday connector are available only for users on the Sana Enterprise tier.
Your Sana workspace owner has enabled this connector – if the workspace owner has disabled the connector it won't appear in the list of available connectors.
Step 1: Go to the connectors page in Sana
In Sana, click on your profile icon and name in the bottom left corner and select Connectors.
Step 2: Find Langfuse in the list of available connectors
Scroll to the Available connectors section and locate Langfuse.
Note: If you can't find a connector in this list, it means one of the following:
You've already connected it — check the ”Active connectors” section at the top of the connectors page, or
Your workspace owner has disabled this connector for your workspace — reach out to them for more information or to request that it be enabled.
Step 3: Hover the connector card and click the connect button
Hover the Langfuse card and click the Connect button.
Tip: If you click the card itself instead of the Connect button, you'll open a connector detail page with more information about its capabilities.
Step 4: Initialize the Pipedream connection
When clicking Connect in the previous step, a pop-up will appear powered by Pipedream. Click the Continue button in the pop-up to initialize the set up.
Note: Sana uses Pipedream to connect your account. Pipedream is a Workday product, trusted by millions of people to keep their data secure.
Step 5: Provide your Langfuse credentials
You'll be asked to paste the credentials for Langfuse so Sana can store them securely and use them to call the connector's APIs on your behalf.
Step 6: Complete the set up
You've successfully connected your Langfuse account. Click Continue to complete the setup and start using the connector.
Known limitations
No trace search: Traces can only be fetched one at a time by exact id. Finding a trace means listing observations or scores first and reading the trace id off those results.
Scores are read-only: Evaluation scores can be listed but not created. Add Feedback writes a free-text comment, not a score.
No prompts or datasets: Langfuse prompt management, datasets, and dataset runs are not available.