Engineers are shipping faster, so we made our CI self-driving to keep up

PostHog’s Developer Experience team built an internal Engineering Analytics product to manage CI maintenance as pull request volume rose from 1,354 in October 2025 to more than 6,500 in October 2026, with a longer-term goal of supporting 10,000 monthly PRs. The system ingests GitHub pull request, workflow run, and job data through PostHog’s warehouse source rather than repeatedly querying the GitHub API, then exposes curated CI, merge queue, and test data through a user interface and MCP tools that agents can query. Self-driving agents investigate signals generated by deterministic detectors for flaky checks, persistent default-branch failures, and workflow duration regressions, and can open remediation pull requests. The approach has produced fixes for a consistently failing WebKit Storybook shard, a flaky Rust test race condition, and an unnecessary dependency installation that reduced a job’s duration from 51 to 33 seconds. PostHog reports shipping more than a dozen improvements through the system, while continuing to improve query performance and add features such as CI Explorer, which visualizes workflow relationships. The company suggests others can create similar automated CI maintenance systems by syncing CI data into a warehouse, connecting agents through MCP, and configuring scouts to identify and address recurring CI issues.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.