Agentflow: Flow Automator - Let Your AI Agents Handle Flow

Flow Automator is a batch queue for Google Flow (labs.google/fx). Write every prompt up front, attach reference images to each one, press Run, and come back to one zip holding every clip and still that Flow rendered. It also includes an optional MCP server, so an AI coding agent such as Claude Code can queue prompts, run the batch and save the finished videos to disk without touching the browser. BATCH PROMPTS - Queue any number of prompts. Paste a whole list and split it into one prompt per row. - Each prompt keeps its own reference images, model, aspect ratio, output count and video length. - Runs the queue unattended: submits, waits for Flow to render, captures, moves on. Set the delay between prompts and how many generations may wait on Flow at once. - Retries a render Flow refuses once, then reports it in plain words. Nothing is silently dropped. - Chain prompts: one prompt's first output becomes the next prompt's reference, so a character sheet becomes a scene and the scene becomes a video. - Image to video, the crop step and Flow's asset picker are driven for you. CAPTURE EVERY CLIP, DOWNLOAD ONE ZIP - Every rendered video and image is captured at Flow's original size the moment it lands, including every take of a multi-output generation, not just the one on screen. - Clips go into a library inside the extension, grouped by Flow project, searchable by name and prompt. - Select what you want and download a single zip. Built in the browser, and archives over 4 GB open correctly. - The Activity tab logs each submit, wait, capture and retry with timestamps, so you know what happened to prompt 14 of 40 and why. OPTIONAL: LET AN AI AGENT RUN THE QUEUE The extension exposes its queue to any Model Context Protocol client through a loopback bridge (127.0.0.1 only, token protected, off by default). The built-in setup guide mints the token and prints a ready command for your client. From there the agent can add prompts, start the batch, wait for it to finish and export the clips to a folder, which turns Flow into one step of a larger pipeline. A command-line front end runs the same commands from a terminal. Requires Node 20 or newer. OTHER OPTIONAL FEATURES, OFF UNTIL YOU TURN THEM ON - Cloud storage: mirror every captured clip to an S3-compatible bucket you own, through a small signing worker you deploy. - Persistent memory: keep Flow project names and clip identifiers across a reinstall or a second machine. - Flow Agent mode: send messages to Flow's built-in agent, manage its instructions and sessions, and approve its proposals from the queue. PRIVACY Prompts, clips and settings stay in your Chrome profile. The extension runs only on labs.google and contains no analytics or ads. Nothing is sent to us unless you enable persistent memory, and then only identifiers, never clips or account details. Full policy: https://flow-automator.pages.dev/privacy REQUIREMENTS A Google Flow account at labs.google/fx. WHO IT IS FOR Creators producing AI video at volume, storyboard and animation teams that need a character to stay consistent across scenes, agencies rendering many variations per brief, and developers who want an agent to drive Flow as part of a pipeline. Google, Google Flow and Veo are trademarks of Google LLC. Flow Automator is an independent project and is not affiliated with, endorsed by or sponsored by Google.