---
type: "post"
id: "langfuse-october-update-langfuse-cbea97"
slug: "langfuse-october-update-langfuse-cbea97"
title: "Langfuse October Update - Langfuse"
description: "Langfuse’s October update adds versioned agent-skill management and tracing alongside scoped, expiring API keys, a redesigned session view, SDK upgrades and evaluator changes."
retrieval_nugget: "Langfuse’s October update adds versioned agent-skill management and tracing alongside scoped, expiring API keys, a redesigned session view, SDK upgrades and evaluator changes. For teams building with agents, it brings skill artifacts into an existing observability workflow and adds more control over what instrumented applications can do with project keys."
published_at: "2026-11-10"
updated_at: "2026-11-10"
record_date: "2026-11-10"
date_kind: "published_at"
topics: ["agent-observability","skills","evals"]
entities: ["Langfuse","OpenAI","gpt-6-luna","TypeSafe","System One","SubQ Snap","Strands Decider"]
related_patterns: ["skills-become-portable-capability-layer"]
novelty_label: "new_fact"
what_changed: "Langfuse’s October update adds agent skill management, a redesigned session view, scoped and expiring API keys, updated SDKs, and Langfuse Government."
prior_refs: []
source_url: "https://langfuse.com/blog/2026-10-31-langfuse-october-update"
source_urls: ["https://langfuse.com/blog/2026-10-31-langfuse-october-update"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:langfuse-october-update-langfuse-cbea97"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/langfuse-october-update-langfuse-cbea97/","markdown":"https://newruntime.com/posts/langfuse-october-update-langfuse-cbea97.md","json":"https://newruntime.com/posts/langfuse-october-update-langfuse-cbea97.json"}
markdown_url: "https://newruntime.com/posts/langfuse-october-update-langfuse-cbea97.md"
---

# Langfuse October Update - Langfuse

## Retrieval answer

Langfuse’s October update adds versioned agent-skill management and tracing alongside scoped, expiring API keys, a redesigned session view, SDK upgrades and evaluator changes. For teams building with agents, it brings skill artifacts into an existing observability workflow and adds more control over what instrumented applications can do with project keys.

## Summary
Langfuse’s October update adds versioned agent-skill management and tracing alongside scoped, expiring API keys, a redesigned session view, SDK upgrades and evaluator changes. For teams building with agents, it brings skill artifacts into an existing observability workflow and adds more control over what instrumented applications can do with project keys.

## What happened
Langfuse’s October 31, 2026 announcement says users can now store and version agent skills alongside prompts. It defines a skill as a `SKILL.md` containing task instructions and optional supporting files. Teams can create skills in the UI, push a local folder, or publish through the API, CLI, MCP and SDKs. Each version preserves the complete set of files. Labels allow a team to promote or roll back a version without changing application code. When skill tracing is configured, generations show which skills they used; a usage tab combines usage counts, file loads and linked observations.

The update also redesigns the session view to show user and assistant messages across traces as one continuous conversation. A sidebar navigates messages, traces and tool usage, with a jump to the trace behind an interaction. Project API keys now have four permission options: Full access, Viewer, Ingestion and Scores ingestion. Langfuse gives examples of an application key that can send traces but cannot read prompts or datasets, and a web-app key that can submit scores only. Keys can have expiration dates; existing keys continue working as before.

Decision-model evaluators now work with OpenAI’s gpt-6-luna through the Decisions API and with models serving TypeSafe’s System One API, including Subquadratic’s SubQ Snap and a self-hosted Strands Decider. Choice, Score and Yes/no questions remain unchanged. The Langfuse MCP server supports these evaluators, while API and CLI support is described as forthcoming.

Langfuse lists Python SDK v5 and JS/TS SDK v6, both built for Langfuse v4. The announcement says most users can upgrade with a version bump because the major releases primarily remove already-deprecated APIs. Changes include experiment comparison links, filterable dotted paths for nested metadata, gzip-compressed span batches and batched score creation. Self-hosted users need Langfuse v4.59.0 or later. The update also lists Claude Platform and Cursor tracing integrations, organization-level LLM connections, usage and export improvements, and outbound host allowlists for self-hosted deployments. Langfuse Cloud becomes v4-only on November 16, 2026; projects with required migration actions should complete them using the Migration Assistant.

## What the archive adds
New Runtime’s earlier coverage on October 21, 2026, described LangChain’s October releases as making the agent harness and evaluation loop an operational product surface ([October 2026 LangChain newsletter](https://newruntime.com/posts/october-2026-langchain-newsletter-everything-we-shipped-at-7e3426/)). This is a new Langfuse product update, not a new fact about that LangChain story; the shared concern is operating and improving agent workflows.

## Trend signal
This update supports the view that skills are becoming a portable capability layer: Langfuse treats skills as versioned files that can be published through several interfaces and traced in use. The fit is partial: the announcement does not establish portability across agent runtimes or address provenance and security review in depth. See [Skills become a portable capability layer](https://newruntime.com/patterns/skills-become-portable-capability-layer/).

## Why now
Langfuse says its Cloud service becomes v4-only on November 16, 2026, and directs users to check the Migration Assistant for required project actions. That near-term cutover makes the SDK and version requirements more than routine release detail for teams on Cloud or self-hosted v4 deployments.

## Why it matters
Versioned skills create a place to manage changes to agent instructions and supporting files without tying every update to an application-code change. Tracing which skills a generation used can help teams investigate behavior and compare outcomes, although the announcement does not claim that it evaluates skill quality automatically.

The API-key scopes offer a practical separation between sending traces and reading prompts or datasets, or between full project access and score submission. Teams can map keys to specific application roles and set expirations rather than sharing a broadly privileged credential. The result is a more inspectable workflow spanning the artifact, its use and the telemetry around it. These capabilities can reduce operational friction, but they do not replace deciding who owns a skill, reviewing changes or validating agent behavior.

## Limits and uncertainty
The announcement describes product capabilities but gives no adoption figures, measured impact on agent quality, or performance and cost comparisons for skill tracing or evaluator choices. It does not quantify the security benefit of scoped keys or explain how skill changes are approved. Langfuse says existing keys continue to work, but teams should verify how their own permissions and expiration policies map to the new options. The stated SDK and migration requirements are vendor guidance, not independently tested compatibility results.

## Skill to build now
Pick one agent workflow whose instructions already live in a `SKILL.md`. Put the file and its supporting assets into Langfuse, label a candidate version, and configure tracing to confirm that generations record the skill in use. Before promoting it, compare the resulting observations with the current version and define who can approve a rollback. This tests the workflow Langfuse describes without assuming that versioning alone improves results.

## Reader takeaway
Langfuse is moving agent skills into the same operational loop as prompts: teams can version them, trace their use and roll back by label. The useful change is not just another place to store instructions; it is the possibility of connecting a specific skill version to observed agent behavior, alongside tighter controls over the keys that send or read that telemetry.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Agent Observability](https://newruntime.com/topics/agent-observability.md): Explore the agent-observability topic hub.
- topic hub: [Skills](https://newruntime.com/topics/skills.md): Explore the skills topic hub.
- topic hub: [Evals](https://newruntime.com/topics/evals.md): Explore the evals topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Sources](https://newruntime.com/sources/): See the original sources this archive cites.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)