[SUBMISSION] FlowForge - A hybrid recorder to supercharge browser-use for Contextual AI Actions and QA.

[shivamkhatri]

Twitter/X Demo Post

https://x.com/ShivamKhatri_/status/1965906282675843116

GitHub Repository

https://github.com/shivamkhatri/flowforge

Project Name

FlowForge - A hybrid recorder to supercharge browser-use for Contextual AI Actions and QA.

Contextual AI Actions

demo.mp4

---

What does it do?

AI copilots today are stateless—they only respond prompt-by-prompt, without understanding the larger workflow. This forces users to repeatedly re-explain context instead of letting AI learn from the full journey.

Problem

• Current copilots can’t observe user actions across steps.
• They react only to a single screen, not the end-to-end process.
• As a result, they can’t recommend meaningful, high-level automations.

Solution: Hybrid Flow Recorder

A Flow Recorder that captures both:

1. Manual actions by the user
2. AI agent-driven actions

This creates a unified sequential log, which a Contextual AI engine uses to recommend relevant next steps. Once approved, the browser-use agent executes them.

Examples

• Context-aware recommendation: While copying tasks into a calendar, AI suggests: “Create calendar events automatically?”
• Proactive completion: While browsing apartments, AI offers: “Compare top 3 apartments” or “Schedule a viewing.”

For QA & Automation

• Flows export as JSON → generate Cypress/Playwright/Selenium tests.
• Reuse flows as manual/agent actions in browser-use.
• Integrates into workflow-use and QA-use for a best-in-class no-code/low-code automation platform.