langfuse-agent 0.2.0

Agent for interacting with Langfuse Observability API

Langfuse Agent - A2A | AG-UI | MCP

Version: 0.2.0

Overview

Langfuse Agent MCP Server + A2A Agent

Agent for interacting with Langfuse Observability API

This repository is actively maintained - Contributions are welcome!

MCP

Available MCP Tools

The MCP server provides 87 tools across 26 categories:

• Annotation Queues: List, create, get, update, delete queues and items
• Blob Storage Integrations: Get, upsert, check status, delete integrations
• Comments: Create, get comments
• Dataset Items: Create, list, get, delete dataset items
• Dataset Run Items: Create, list dataset run items
• Datasets: List, create, get, delete runs, get runs
• Health: Health check
• Ingestion: Batch ingestion
• Legacy Metrics: V1 metrics
• Legacy Observations: V1 get observations
• Legacy Score: V1 create/delete scores
• LLM Connections: List, upsert connections
• Media: Get, patch, get upload URL
• Metrics: Metrics query
• Models: Create, list, get, delete models
• Observations: Get many observations
• OpenTelemetry: Export traces
• Organizations: Manage organization and project memberships
• Projects: Get, create, update, delete projects and API keys
• Prompts: Version update, get, delete, list, create prompts
• SCIM: Service provider config, resource types, schemas, user management
• Score Configs: Create, get, update score configs
• Scores: Get many scores, get by ID
• Sessions: List, get sessions
• Traces: Get, delete, list, delete multiple traces

Using as an MCP Server

The MCP Server can be run in two modes: stdio (for local testing) or http (for networked access).

Environment Variables

• LANGFUSE_URL: The URL of the target service.
• LANGFUSE_TOKEN: The API token or access token.

Run in stdio mode (default):

export LANGFUSE_URL="http://localhost:8080"
export LANGFUSE_TOKEN="your_token"
langfuse-mcp --transport "stdio"

Run in HTTP mode:

export LANGFUSE_URL="http://localhost:8080"
export LANGFUSE_TOKEN="your_token"
langfuse-mcp --transport "http" --host "0.0.0.0" --port "8000"

A2A Agent

Run A2A Server

export LANGFUSE_URL="http://localhost:8080"
export LANGFUSE_TOKEN="your_token"
langfuse-agent --provider openai --model-id gpt-4o --api-key sk-...

The A2A agent uses agent-utilities for:

• Auto-discovery of MCP tools from mcp_config.json
• Pydantic AI agent with graph-based routing
• Web UI support
• OpenTelemetry integration
• Custom skills support

Docker

Build

docker build -t langfuse-agent .

Run MCP Server

docker run -d \
  --name langfuse-agent \
  -p 8000:8000 \
  -e TRANSPORT=http \
  -e LANGFUSE_URL="http://your-service:8080" \
  -e LANGFUSE_TOKEN="your_token" \
  knucklessg1/langfuse-agent:latest

Deploy with Docker Compose

services:
  langfuse-agent:
    image: knucklessg1/langfuse-agent:latest
    environment:
      - HOST=0.0.0.0
      - PORT=8000
      - TRANSPORT=http
      - LANGFUSE_URL=http://your-service:8080
      - LANGFUSE_TOKEN=your_token
    ports:
      - 8000:8000

Configure mcp.json for AI Integration (e.g. Claude Desktop)

{
  "mcpServers": {
    "langfuse": {
      "command": "uv",
      "args": [
        "run",
        "--with",
        "langfuse-agent",
        "langfuse-mcp"
      ],
      "env": {
        "LANGFUSE_URL": "http://your-service:8080",
        "LANGFUSE_TOKEN": "your_token"
      }
    }
  }
}

Install Python Package

python -m pip install langfuse-agent

uv pip install langfuse-agent

Repository Owners