Decipher AI (getdecipher.com) is a YC W24 company founded in 2023 by Michael Rosenfield and Rohan Das, both ex-Google engineers who led the YouTube Homepage and Pixel product lines. The company started as an AI-powered session replay analysis tool that used vision-language models (VLMs) to watch user sessions and surface bugs, feature-usage patterns, and UX friction points. It has since evolved into an agentic QA platform squarely targeting teams that ship with AI coding agents. The current core product is an AI QA agent that observes your product, generates end-to-end (E2E) tests in minutes, and maintains those tests automatically as the UI changes. When a test fails, Decipher distinguishes between a real bug, a changed flow, or a flaky dependency, so engineers spend time on genuine issues rather than broken selectors. The platform integrates directly with Claude Code, allowing Claude to generate, run, and maintain E2E tests through Decipher's infrastructure without leaving the coding workflow. Decipher also runs continuous production monitoring: its AI agents watch live customer sessions to build a map of real user journeys, golden paths, and edge cases, and fire an alert with session replay, console/network context, and customer-impact data the moment a user hits a bug. This closes the loop between pre-production testing and post-production observability. Customers include Bilt, Arize, and Vial. The company is based in San Francisco and has raised a pre-seed round from Y Combinator and BoxGroup (approximately $500K YC standard terms, April 2024). The product is positioned as a validation layer for the age of AI-accelerated development, where human QA cannot keep pace with the volume of code that coding agents produce. Key features: - AI agent generates E2E tests from watching your product or reading natural-language prompts, no manual scripting - Self-healing tests: automatically updates test steps when UI or product flows change - Intelligent failure analysis: classifies failures as real bug, flow change, or dependency issue - Production session monitoring via vision-language models to detect live bugs with replay + console/network context - Claude Code integration: Claude can generate and run self-maintaining E2E tests through Decipher's infrastructure - Customer deep-research: ask plain-English questions about a specific account and get answers derived from session data - Smart Events: define new analytics events in under 15 seconds with natural language, no engineering required - SOC 2 Type II audited; client-side data masking before data reaches servers
Not public (pricing page exists at getdecipher.com/#pricing but tiers were not confirmed by official source in search results; third-party sources cite a free tier and paid plans from ~$5 to $169/month, unverified)