Langfuse Service#

The Langfuse service provides LLM observability and evaluation workflows with 37 tools for traces, sessions, prompts, scores, datasets, models, annotation queues, metrics, project management, membership management, and API key management.

Overview#

The Langfuse service in Cloud Native MCP Server lets AI assistants inspect prompt executions, user sessions, observations, and evaluation results from Langfuse using the Public API. It is useful when you want to trace how an LLM interaction behaved end to end without leaving the MCP surface.

Key Capabilities#

### 🧭 Trace Discovery Browse traces and drill into a single trace when you need full detail.

### 🧵 Session Correlation Inspect sessions to connect multiple traces in one user flow.

### 📝 Prompt Intelligence Look up prompt versions, labels, and resolved prompt content.

### 📏 Evaluation Insights Inspect scores and query Langfuse metrics for trend analysis.

Available Tools (37)#

Health#

langfuse_check_health: Check Langfuse API and database health

Traces#

langfuse_list_traces_summary: Compact trace discovery view
langfuse_list_traces: Full trace listing with filters
langfuse_get_trace: Get a specific trace by ID

Annotation Queues and Datasets#

langfuse_list_annotation_queues: List annotation queues
langfuse_get_annotation_queue: Get one annotation queue
langfuse_list_annotation_queue_items: List queue items
langfuse_list_datasets: List datasets
langfuse_get_dataset: Get one dataset
langfuse_list_dataset_runs: List dataset runs
langfuse_get_dataset_run: Get one dataset run

Sessions and Observations#

langfuse_list_sessions: List sessions
langfuse_get_session: Get a specific session
langfuse_list_observations: List observations
langfuse_get_observation: Get a specific observation

Models and Score Configurations#

langfuse_list_llm_connections: List LLM connections
langfuse_list_models: List models
langfuse_get_model: Get one model
langfuse_list_score_configs: List score configurations
langfuse_get_score_config: Get one score configuration

Prompts, Scores, and Metrics#

langfuse_list_prompts: List prompt versions
langfuse_get_prompt: Get a prompt by name
langfuse_list_scores: List scores and evaluations
langfuse_get_score: Get a specific score
langfuse_get_metrics: Run a Langfuse metrics query

Projects#

langfuse_get_project: Get the project associated with the configured credentials
langfuse_list_organization_projects: List organization projects
langfuse_create_project: Create a project
langfuse_update_project: Update a project
langfuse_delete_project: Delete a project
langfuse_list_project_memberships: List project memberships
langfuse_upsert_project_membership: Create or update a project membership
langfuse_delete_project_membership: Delete a project membership

API Key Management#

langfuse_list_organization_api_keys: List organization API keys
langfuse_list_project_api_keys: List project API keys
langfuse_create_project_api_key: Create a project API key
langfuse_delete_project_api_key: Delete a project API key

Configuration Example#

1
2
3
4
5
6
langfuse:
  enabled: true
  url: "https://cloud.langfuse.com"
  username: "pk-lf-..."
  password: "sk-lf-..."
  timeoutSec: 30

Langfuse uses HTTP Basic Auth on the Public API:

Username: configured Langfuse Basic Auth username
Password: configured Langfuse Basic Auth password

The project, membership, and API key management tools require organization-scoped Langfuse credentials.

Next Steps#

OpenTelemetry Service for infrastructure telemetry
Jaeger Service for distributed trace troubleshooting
Configuration Guide for environment variables and endpoints