Source: github.com/langfuse/langfuse/releases/tag/v4.83.0
The AI gateway gains per-key spend limits and a new Budgets page for per-user spend budgets, and default spend budgets can now be set for keys and users. Bedrock model connections can use IAM role authentication.
The MCP server now serves the 2026-07-28 revision of the Model Context Protocol and has a batchCreateScores tool. The server has also been migrated to version 2 of the TypeScript SDK.
The release includes many fixes and maintenance changes across the web app, CI and billing, and it adds default pricing for gpt-6.2-astra, gpt-6.2-sol and several gemini-4 models. The changelog lists the full set of changes from v4.82.0, and 18 contributors are credited.
Why it matters if you build agents
If your agents call Langfuse's MCP server, the server now speaks the 2026-07-28 protocol revision and can write scores in batches with batchCreateScores.
If your agents' model calls pass through the Langfuse AI gateway, you can cap spend per key and per user rather than only reviewing usage afterwards.
If you connect Bedrock through Langfuse, check whether IAM role authentication fits your setup before you rely on other credentials.
Get stories like this every weekday
One short email with the day's agent tooling news.