Agentic access to observability
AI agents can investigate production behavior directly in Langfuse. There are different ways for agents to access your data:
Choose an access method
| Agent capabilities | Recommended access |
|---|---|
| Can install tools and run shell commands | Install the Langfuse Agent Skill. It teaches the agent Langfuse workflows and uses the Langfuse CLI to query and update project data. |
| Cannot install tools or run shell commands | Connect the Langfuse MCP server to expose Langfuse operations as tools. |
| Runs as part of a script or CI/CD pipeline | Use the Langfuse CLI directly or call the Public API. |
Example workflows
Ask your agent to:
- Find observations with errors or high latency in a specific environment
- Compare token usage, cost, or latency across models and releases
- Inspect the input, output, metadata, and scores of a problematic observation
- Add a comment or score to record the outcome of an investigation
Work across Langfuse
Agents can also manage prompts and run evaluation workflows in Langfuse.
Related guides and blog posts
- Headless Langfuse from your coding agent
Instrument an application, analyze production traces, build a dataset, and run evaluations without leaving your coding agent.
- How we use agents to review production infrastructure
See how repo-owned skills, MCP servers, and queryable production data support recurring engineering reviews.
Was this page helpful?
Last edited
