This blog post details how the OpenAI Developer Community used Discourse Workflows and an AI agent to synchronize a single topic with a daily X feed. Following Tibo’s “28 days of shipping” announcement series, co-founder Sam Saffron implemented a solution where Discourse automatically polls X for new posts and updates a central table with summaries, links, and original post references. This eliminates the need for manual copying and allows developers to discuss releases in a threaded format rather than fragmented X replies.
The implementation was built using an AI agent (GPT-6.1 Sol) via Discourse Workflows. The process involved creating a clean environment, defining the goal, and instructing the agent to learn the X API and build a fake version for testing. The agent produced a working workflow with tests, which was then refined to optimize storage and configuration. After exporting the workflow to the OpenAI community, the team used the agent to fix production gaps, as the export file did not initially include data table or agent definitions.
The full build is documented on Meta, designed to be followed by AI agents. Users can prompt their agents to synchronize social media discussions into Discourse topics using the provided method. This approach enables communities to keep long-running topics updated during events, open shadow topics for specific social media activity, and aggregate posts from multiple accounts into searchable records. By leveraging Discourse Workflows, organizations can continue posting on social media where their audience is active while maintaining a lasting, discussion-ready record on their Discourse instance without manual staff intervention.
How we used Discourse Workflows and an AI agent to turn OpenAI's 28 days of shipping announcements on X into a single table on the OpenAI Developer Community