AI Agent News/stories/langfuse-v4-70-0-adds-read-only-topics-endpoints-and-a-batch-evaluation-mcp-tool.md

Infrastructure

Langfuse v4.70.0 adds read-only Topics endpoints and a batch evaluation MCP tool

Langfuse has released v4.70.0, adding unstable read-only Topics endpoints to its public API and a runBatchEvaluation MCP tool. The release also includes AI Gateway setup changes, security fixes for webhook and PostHog destinations, and several web updates.

Source: github.com/langfuse/langfuse/releases/tag/v4.70.0

The release adds unstable, read-only Topics endpoints to the public API, according to the changelog on GitHub. It also adds a runBatchEvaluation MCP tool and a batch_action_id field on batch run outputs.

On the AI Gateway side, the Claude Desktop setup steps were updated and the Claude tabs merged. Gateway keys are now described inline on the keys page, which also lets users invite people. Web changes include syntax highlighting for shell, TOML and JSON setup snippets, and Cmd+K shortcuts that jump to prompts, datasets, queues and evaluators.

Fixes include requiring secrets again when a webhook or PostHog destination changes, plus upgrades to dompurify 3.4.17 and nodemailer 10.1.7. A ClickHouse change labels client-closed streams, and a new warning appears when ClickHouse staff view data.

Why it matters if you build agents

Teams building agents with Langfuse can read Topics data through the API, but the endpoints are unstable and read-only, so they should not depend on them as stable interfaces yet.

Teams that run evaluations from MCP clients get a direct batch evaluation tool.

Anyone with webhook or PostHog destinations should confirm their secrets are set again after changing those settings.

Get stories like this every weekday

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

Subscribe