Start free

Connect Langfuse to your agent for live trace checks, prompts and eval scores.

Langfuse is an open source platform for debugging and improving LLM applications. Connect it to AgentCalls and your agent can read the state of your AI product during a call. It can look up the prompt version you are running, check recent observations and cost, list datasets, and attach a human score to a trace. Your team stops digging through dashboards to answer a question from a customer or a colleague on the phone.

Start freeSee pricing

What your agent can do with Langfuse

  • Fetch the live version of a named prompt by label, such as production or staging.
  • Look up recent LLM observations, tool calls and their latency or cost.
  • List datasets and experiments so your team can answer status questions on the phone.
  • Attach an evaluation score to a trace, session or dataset run after a call.
  • Check which annotation queues are waiting for human review.

Langfuse exposes 10 actions. You choose which ones the agent may use and whether each runs during or after the call.

Example calls

Confirm the prompt version in production

  1. Caller

    Hi, this is Dana from the product team. Which prompt version is live for the support bot right now?

  2. Agent

    Sure, Dana. The support-triage prompt is on the production label, version 14, last updated on the third of June.

  3. Caller

    Good. Can you check whether the staging version has a higher version number?

  4. Agent

    Staging is on version 15. It has not been promoted to production yet.

Check cost and latency after a complaint

  1. Caller

    A customer said the assistant was slow yesterday afternoon. Do we have data on that?

  2. Agent

    I can see the observations from yesterday between two and four in the afternoon. The median response time was 3.2 seconds, up from 1.9 seconds the day before.

  3. Caller

    What drove the increase?

  4. Agent

    Most of the extra time was in one tool call to the order lookup. I can leave a note on the trace for the engineering team to review.

Score a reviewed conversation

  1. Caller

    I finished reviewing trace 8f2c from this morning. The answer about refunds was correct but too long.

  2. Agent

    Got it. Should I record a score of 3 out of 5 for quality, with that note attached?

  3. Caller

    Yes, record it as a 3, and add the note about length.

  4. Agent

    Done. The score is attached to trace 8f2c with your note.

How to connect Langfuse

  1. 1

    Create your agent

    Sign up to AgentCalls, pick an English or Arabic voice and write what the agent should know and do in plain language.

  2. 2

    Connect Langfuse

    From Integrations in the dashboard, choose Langfuse and enter the connection details from your Langfuse account settings.

  3. 3

    Choose the actions

    In the agent editor, enable the Langfuse actions the agent may use and set each to run during the call or afterwards. Then test it from your browser.

Related integrations

Perplexity AICustomer supportPerplexity AI provides conversational AI models for generating human-like text responsesSerpApiSpreadsheets & databasesSerpApi provides a real-time API for structured search engine results, allowing developers to scrape, parse, and…FirecrawlAI, search & researchFirecrawl automates web crawling and data extraction, enabling organizations to gather content, index sites, and gain…TavilyAI, search & researchTavily offers search and data retrieval solutions, helping teams quickly locate and filter relevant information from…ExaSpreadsheets & databasesExa focuses on data extraction and search, helping teams gather, analyze, and visualize information from websites…ElevenLabsAI, search & researchCreate natural AI voices instantly in any language - perfect for video creators, developers, and businesses.

All AI, search & research integrations

Questions, answered.

Does the agent work with Langfuse during the call or only after it?

Both. During a call your agent can look up prompts, traces, datasets and experiments. It can also attach a score to a trace once the call is over or when you ask it to.

What do I need to connect Langfuse?

You need a Langfuse project and its public and secret keys, which you enter from the Langfuse settings. Then sign in from the AgentCalls dashboard and connect the tool. Integrations are included from the Growth plan.

Can the agent change my prompts or delete data?

The agent can read prompts and traces and add evaluation scores. It does not edit prompts or remove data through this integration.

Page drafted from the tool's public description and reviewed automatically for accuracy of claims

Put your agent to work with Langfuse.

Connect in minutes, test from your browser, go live on a number in the US, UK or Gulf. Free to start, no card needed.

Start freeSee pricing