AI Agent News/stories/langfuse-v4-63-0-separates-ai-gateway-keys-and-expands-skill-management.md

Infrastructure

Langfuse v4.63.0 separates AI Gateway keys and expands skill management

Langfuse has released v4.63.0, a routine update with new features across its AI Gateway, skills and in-app agent. The release also includes a set of bug fixes and dependency patches.

Source: github.com/langfuse/langfuse/releases/tag/v4.63.0

The release separates AI Gateway keys from organization API keys, according to the changelog, and validates connection names on the server with autofill. Gateway errors are now attributed by origin, and the team says it can measure gateway overhead. Requests for slugless codex- models are routed to OpenAI connections.

Skills gain a usage metrics tab and a public v4 preview of skill management. The in-app agent can now ask the user a short questionnaire. Annotation queues make score configuration setup visible and ordered.

Among the fixes, handled MCP timeouts are no longer classified as unexpected errors. The release also removes the LANGFUSE_SSO_DISCOVERY_ALLOWED_HOSTS variable and counts dual-write observations in the billable units ledger. Dependencies qs, nodemailer and hono were patched.

Why it matters if you build agents

Teams that route model calls through the AI Gateway can now use gateway keys that are separate from organization API keys. Errors and gateway overhead are reported with clearer attribution, which helps when debugging agent traffic.

Self-hosted deployments that set LANGFUSE_SSO_DISCOVERY_ALLOWED_HOSTS should check their configuration after upgrading, since the release removes that variable.

Get stories like this every weekday

One short email with the day's agent tooling news.

Subscribe