Temporal and Braintrust have integrated to solve durability and observability challenges in AI agent development. The integration combines Temporal's durable workflow execution (automatic retries, state persistence) with Braintrust's LLM observability (tracing, prompt versioning, evals, cost metrics). A practical deep research agent example demonstrates how to set up the integration with just a few lines of code: wrapping the OpenAI client, adding the Braintrust plugin to Temporal workers, and managing prompts through Braintrust's UI. This enables developers to build production-ready AI agents that survive failures, provide full visibility into LLM calls, and allow prompt iteration without code deployments.
Table of contents
The problem: AI agents in the real world #What is Temporal? #What is Braintrust? #The integration: Better together #Demo: Deep research agent #What you can see in Braintrust #What you can see in Temporal #Try it yourself #Sort: