The LinkedIn Autopilot
A daily LinkedIn system that reads the AI news, throws out anything covered before, and writes one post in your voice. You get a Telegram card with Approve or Skip, so it takes ten seconds a day and nothing publishes without your thumb. Built on n8n, Claude, Telegram, and Blotato.
Section One
What You're Building
A system that keeps you visible on LinkedIn every single day for about ten seconds of your attention. Every morning at 7am it reads your chosen AI news sources, filters to the last 36 hours, throws out anything it has covered before, and hands the survivors to Claude. Claude picks the one story that actually matters to business owners, writes the post in your voice, and sends the draft to your phone.
You get a Telegram card with two buttons. Approve publishes to LinkedIn through Blotato and banks the post into your content library for repurposing. Skip blacklists the story forever. Ignore it, and the draft quietly expires by evening. Nothing ever posts without your thumb on the button.
Fig. 1 — system flow
Source
RSS + YouTube
Filtered to the last 36 hours
Dedupe
Post log
Never repeats a story
Write
Claude
Picks 1 story, writes the post in your voice
Approve
Telegram → Blotato
Approve/Skip, then publishes to LinkedIn
The rule that makes this work
The approval gate is the design, not a training wheel. Consistency is the whole game on LinkedIn, and this makes the default outcome a good post instead of no post, while the Telegram buttons keep your judgment in the loop at a cost of ten seconds a day.
Section Two
What You Need
n8n
The automation engine and the thing Claude builds inside. Cloud starter works fine; this workflow uses one execution per day. You also need a Claude plan that supports connectors, since the whole build method is Claude driving n8n directly.
Claude API key
Powers story selection and writing. A month of daily posts costs less than one coffee on Sonnet.
A Telegram bot
Your approval remote control. Message @BotFather, send /newbot, name it, and save the API token it gives you. Then open your new bot and send it /start so it can message you back.
Blotato
The publishing bridge to LinkedIn. Connect your LinkedIn account in the Blotato dashboard and note the account ID it assigns. Grab your API key from settings.
Airtable
The content bank. Every published post lands in a table so you can repurpose it into reels, carousels, and newsletters later. Skip it and the system still runs.
Total stack cost if you're starting from zero: roughly $25 to $40 a month, most of it n8n. If you already run n8n for anything else, the marginal cost of this system is pocket change.
Section Three
Step-by-Step Build
01
Step One
Connect Claude to n8n
In n8n: Settings, then n8n API, create an API key. In Claude (claude.ai): Settings, Connectors, Add custom connector, name it n8n, and use your instance MCP URL: https://YOUR-INSTANCE.app.n8n.cloud/mcp-server/http with the API key. Once connected, Claude can create, edit, run, and debug workflows in your account.
02
Step Two
Two-minute prerequisites
Telegram: message @BotFather, send /newbot, save the token, then open your new bot and send it /start so it can message you back. Get your own chat ID by messaging @userinfobot. Blotato: connect LinkedIn in the Blotato dashboard and note the account ID plus your API key from settings. Add both credentials in n8n (Telegram API, and a Header Auth credential named blotato-api-key).
03
Step Three
Paste the build prompt
Open a Claude chat with the n8n connector on, paste the prompt below with your three values filled in, then paste the system prompt from the next step when Claude asks for it. Then watch it work. A full build takes a few minutes.
Copy this: the build prompt
Connect to my n8n and build a workflow called LinkedIn AI News Autopilot. My values: Telegram chat ID = [YOURS], Blotato LinkedIn account ID = [YOURS], timezone = [YOURS]. The system, in order: 1. Schedule trigger, daily 7:00am. 2. A source list node with my feeds (one item per feedUrl): - [RSS FEED URL 1: your industry news source] - [RSS FEED URL 2] - [RSS FEED URL 3] - https://www.youtube.com/feeds/videos.xml?channel_id=[CHANNEL_ID] (any YouTube channel has this hidden feed) 3. RSS Read per feed, continue on error so a dead feed never kills the run. 4. Normalize items to url, title, snippet, source, publishedAt. Keep only the last 36 hours, dedupe, cap at 25. 5. Create a data table named linkedin_content_log with columns url, title, source, status, postedAt. Pass through only stories whose url is NOT already in the table. 6. A Claude (Anthropic) node that runs once with ALL candidates and uses the system prompt I will provide. It must return strict JSON: chosen_url, chosen_title, source, post_text. 7. A code node that parses that JSON and fails loudly if the reply is not valid JSON. 8. Telegram send-and-wait approval to my chat ID with buttons Post it / Skip, 12 hour wait limit. 9. On approve: HTTP POST to https://backend.blotato.com/v2/posts with my blotato-api-key header credential, publishing the post text to my LinkedIn account ID. Then log the story to the data table with status posted. 10. On skip or timeout: log it with status skipped so it never comes back. Use my existing n8n credentials for Telegram, Anthropic, and Blotato. Remember every dynamic value in n8n needs the = expression prefix. Build it, then validate it, and tell me what you built before running anything.
Finding feeds for any niche
Almost every news site publishes RSS: try adding /feed to the homepage URL, or ask Claude to "find the RSS feeds for the top 5 news sources in [your industry]". For YouTube, open any channel page, view source, and search channelId. Three to five sources you respect beat twenty generic ones, because the system can only pick from what you feed it.
04
Step Four
Paste the system prompt into the Claude node
This is the asset. The build creates the Claude node empty; this prompt is what makes the output sound like you instead of every other AI-written post. Replace the bracketed lines with your name, background, business, and audience. Keep the structure rules and the bans exactly as written.
Copy this: system prompt
You write LinkedIn posts for [NAME], a [YOUR BACKGROUND] who runs [YOUR BUSINESS], serving [YOUR AUDIENCE]. Voice: direct, practical, zero hype, no emojis, no buzzwords like game-changer or revolutionary. Translate AI news into what it actually means for a business owner running operations. Post structure: one-line hook stating the concrete development, two or three short sentences on what happened, then an operator take on why it matters for people running real businesses, ending with one genuine question to invite comments. 120 to 180 words. No more than 3 hashtags, placed at the very end. Prefer stories about AI agents, workflow automation, and practical business AI over model benchmark news or funding rounds. Never invent facts not present in the snippet.
Why the bans matter
"Never invent facts not present in the snippet" is the trust anchor — the difference between a system you approve with one tap and one you fact-check every morning, which defeats the point. The no-hype and no-emoji rules are what keep it from sounding generated.
Reference: the message sent each run
Here are today's candidate stories as JSON: [...candidates...]
Pick the ONE story most relevant to [YOUR AUDIENCE]. Then write the LinkedIn post. Return ONLY valid JSON, no markdown fences, exactly this shape:
{"chosen_url":"...", "chosen_title":"...", "source":"...", "post_text":"..."}05
Step Five
Test it with Claude
Paste the test prompt below. Claude runs the workflow, reads its own execution logs, and repairs whatever breaks. This loop is the entire debugging experience: you never open an error message yourself.
Copy this: the test prompt
Run the workflow manually now. Check the execution: if any node failed, read the error, fix the node, and run it again. Repeat until the Telegram approval card arrives on my phone. Do not activate the workflow yet. When the run reaches waiting status, summarize what happened at each node.
Why this works
The approval card arriving on your phone IS the test passing. Everything upstream of Telegram had to work for that card to exist, so one visible result verifies nine invisible steps.
06
Step Six
Approve once, then activate
Tap Post it on the test card and confirm the post appears on LinkedIn. Then tell Claude: "Activate the workflow." From tomorrow at 7am it runs itself, and your entire job is one Telegram tap a day.
Good to know
Your Blotato account ID lives in the Blotato dashboard under your connected LinkedIn account, and the API key is in settings. If the publish step returns an auth error, the key belongs in a header named blotato-api-key, not in the request body.
Section Four
The Upgrade Path
Run it with the approval gate for two or three weeks. You're not just checking for errors; you're teaching yourself what the prompt gets wrong so you can tighten the voice. Once you've approved twenty drafts in a row without edits, deleting the Telegram node makes the system fully autonomous. That's a one-minute change, and it should be earned, not assumed.
From there the same skeleton scales sideways. Blotato publishes to Twitter/X, Facebook, Threads, and more, so one extra branch turns daily LinkedIn into daily everywhere. The Airtable bank feeds a second workflow that turns your best posts into reel scripts and carousels. And the source list is a living asset: prune feeds that never win, add ones that do.
“The approval card arriving on your phone IS the test passing. Everything upstream of Telegram had to work for that card to exist.”
— The LinkedIn Autopilot, p. 7
§ Fin.
Rather have it built for you?
This guide is one system from the Soulwire Studio workflow library. We build AI automations for businesses that run on process: intake, scheduling, proposals, compliance-heavy operations. If you'd rather spend your time closing deals than debugging content, book an audit and we'll find the automation hiding in your business.
