Skip to content
Founder Workflows
BrowseSaved
Browse by stage

Publish a daily sourced news post to a platform with no API using an unattended browser agent

An open-source agent researches the last day's news, verifies sources, renders image cards and publishes a daily Xiaohongshu note by driving a logged-in browser. Over 36 days it published 28 of 29 scheduled runs.

DocumentedDistributionContent and newsletterScheduled automationScript

Evidence: The source shows its work: steps, screenshots, code or data you can inspect.

The business problem

Content creators want a consistent daily post on platforms that offer no publishing API, but doing the research, design and posting by hand every day is slow, and simple automations break or post duplicates.

What was tried

1. Keep a dedicated Microsoft Edge profile logged in to the platform and start it in debug mode. 2. Each morning a cron job (09:00) starts the agent, which runs 3 to 4 keyword searches over the past 24 to 48 hours with web search and extraction tools, opens each source to confirm it, and drops anything it cannot verify. 3. The agent writes a spec file with five key points and their source links. 4. Scripts build the cover, point cards and body text from the spec and render them as 1080x1440 images from HTML. 5. The browser is driven over the Chrome DevTools Protocol (no Playwright or Selenium) to upload the images, fill the title and body in the editor, and press publish. 6. Three checks confirm success: the page address, the review-to-published status, and the absence of a permission message. The repository documents 20 failure modes in a separate file.

What was reported (not reported)

Reliability only. Between 2026-09-14 and 2026-10-19 (36 days) there were 29 scheduled runs, 28 successful publishes and 1 failure caused by a model API timeout after 601 seconds, with no duplicate posts. Each run took 4 to 8 minutes end to end. No followers, views, engagement or revenue are reported, and content quality is explicitly out of scope.

Limitations

Personal experiment with one commit and very few stars; no audience or business results. The language model used is not named. Logins must be renewed by hand when cookies expire, topic tags do not become links, and the scripts depend on page selectors that break when the platform changes. Automating a platform without an official API may break its terms: the author warns against bulk or spam use. Xiaohongshu is a Chinese platform, so the setup suits creators targeting that audience; the pattern transfers to other closed sites only with the same risks.

What you need

Python, a desktop browser kept logged in, a model API with web search, and a machine that stays on for the daily schedule.

Sources

Source published: unknown. Last reviewed here: October 28, 2026. Spot a mistake? Tell us.

Tools in this workflow

Get new workflows like this by email

A short selection of the most useful new workflows, including the ones that did not work. Free, and you can unsubscribe at any time.

Your email address

Related workflows