孫昊

Posted on May 6

How I Programmatically Publish to Substack via Chrome DevTools Protocol

#substack #automation #cdp #playwright

TL;DR: Substack has no public write API in 2026, but its TipTap editor accepts CDP automation reliably. Here's the 50-line Python script that publishes any markdown file to Substack as a fully formatted post in 45 seconds. Saves 5+ minutes per post for a publishing batch of 30+.

Why CDP, not API?

Substack offers a public read API (RSS, API for subscriber list) but no public write API. Their internal API (/api/v1/posts) requires session cookies and CSRF tokens that change per session.

CDP attaches to a real Chrome you've already logged into. No reverse-engineering, no broken sessions — your browser does what it always does.

Setup (one-time, 60 sec)

chrome.exe ^
  --remote-debugging-port=9222 ^
  --remote-allow-origins=* ^
  --user-data-dir="C:\Users\you\AppData\Local\Google\Chrome\User Data\AutomationProfile"

Open this profile, log into Substack manually once. The profile stays logged in indefinitely.

The 50-line Python publisher

from playwright.sync_api import sync_playwright
import re, time, pyperclip
from pathlib import Path

def publish(md_path, target_substack_publish_url):
    raw = Path(md_path).read_text(encoding='utf-8-sig')
    # Strip YAML frontmatter
    raw = re.sub(r'^---\s*\n.*?\n---\s*\n', '', raw, flags=re.DOTALL)
    title_match = re.search(r'^# (.+)$', raw, re.MULTILINE)
    title = title_match.group(1) if title_match else 'Untitled'
    body = re.sub(r'^# .+\n', '', raw, count=1, flags=re.MULTILINE).strip()

    with sync_playwright() as p:
        browser = p.chromium.connect_over_cdp("http://127.0.0.1:9222")
        ctx = browser.contexts[0]
        page = ctx.new_page()
        page.goto(f"{target_substack_publish_url}/publish/posts?type=newsletter")
        page.wait_for_load_state('networkidle')

        page.click("text=新規投稿")  # or "New post" in EN
        page.wait_for_selector('textarea[aria-label="title"]', timeout=10_000)

        page.fill('textarea[aria-label="title"]', title)
        time.sleep(1)

        editor = page.locator("div.ProseMirror").first
        editor.click()
        time.sleep(0.5)
        pyperclip.copy(body)
        page.keyboard.press("Control+V")
        time.sleep(2)

        # Wait for autosave indicator
        page.wait_for_selector('text=下書き保存しました', timeout=15_000)

        page.click("text=続ける")
        page.wait_for_selector("text=今すぐ全員に送信", timeout=10_000)
        page.click("text=今すぐ全員に送信")
        page.wait_for_url(re.compile(r'.*/p/[a-z0-9-]+'), timeout=20_000)

        published_url = page.url
        return published_url

# Usage
url = publish("reports/substack-issue-23.md", "https://autoappnotes.substack.com")
print(f"Published: {url}")

What this saves you

Per issue manual flow:

Open Substack publish UI: 5 sec
Copy title manually: 5 sec
Open editor, paste body: 30 sec
Wait for autosave: 10 sec
Click 続ける: 2 sec
Click 今すぐ全員に送信: 3 sec
Confirm: 5 sec

Total manual: ~60 sec per issue, but with attention required throughout.

CDP automated: 45 sec per issue, fully unattended — you can batch 10 issues from CSV in 8 minutes.

Common errors and fixes

"下書き保存しました" never appears

Substack lost focus on the editor. Click into ProseMirror again first.

Title accepted but body cut off

Body has frontmatter leak. Strip with re.sub(r'^---\s*\n.*?\n---\s*\n', '', body, flags=re.DOTALL).

Ctrl+V pastes nothing

Editor not focused. Use editor.click() before pyperclip.copy().

"今すぐ全員に送信" button missing

Substack has multiple publish modal variants. Use page.click('button:has-text("今すぐ"):visible') for fallback.

Markdown not rendering as headings/lists

TipTap doesn't auto-parse markdown on paste. Either:
- (a) Convert markdown → HTML first, paste HTML
- (b) Convert markdown → ProseMirror JSON, dispatch via editor.evaluate()
- For most cases, (a) with markdown-it works:
```
from markdown_it import MarkdownIt
md = MarkdownIt()
html = md.render(body)
pyperclip.copy(html)  # TipTap accepts HTML clipboard via Ctrl+V
```

Why this scales

Once the script works, you can drive your full content backlog:

for f in reports/substack-issue-*.md; do
    python publish.py "$f"
    sleep 60  # rate limit
done

In one evening I went from "writing in advance" to "publishing 4 weeks of content in 30 minutes."

Source

Full Substack TipTap publish script + ProseMirror JSON converter (for richer formatting):

AutoApp Dashboard ($39) includes:

substack_publish.py (50 lines, this article)
substack_publish_with_html.py (markdown→HTML→TipTap)
substack_publish_batch.py (CSV-driven batch)
substack_subscriber_export.py (read API helper)

If you're publishing to Substack regularly and not automating, you're spending 5 hours per month on copy-paste. CDP solves it in 50 lines.

DEV Community