Trigger, Generate, and Auto-Publish Social Posts via Telegram and OpenAI

Input: Telegram text message → Process: read Google Docs brand rules, OpenAI Agent drafts per platform, Telegram approval → Output: PostWire publishes to LinkedIn/Facebook/Bluesky/Mastodon, results logged to Google Sheets

Who

Content operators or small teams who need brand-consistent multi-platform social publishing

You get

Send a topic via Telegram to generate per-platform drafts; auto-publish after approval with results logged

You need

n8n · telegram · google-sheets · google-docs · serpapi · postwire

Steps

1. Configure Telegram trigger and filter

   Add a Telegram Trigger node in n8n, select your Telegram Bot credential, and listen for message updates. Add a Filter node (Only text topics) with two conditions: message text is not empty, and does not start with “/” (to exclude commands). Once saved, sending a text message to your bot will trigger the workflow.

2. Set up Google Docs brand rules retrieval

   Add a Google Docs node (Get the brand rules), connect your Google credential, and enter your brand rules document URL in the documentURL field. The document content is passed as part of the system context to the AI Agent to control tone, audience, and words to avoid.

3. Configure AI Agent to generate multi-platform drafts

   Add an AI Agent node (Write the posts with an AI agent), set promptType to define. In the text field, use an expression to concatenate the Topic and Brand rules; in options.systemMessage, enter the system instruction requiring JSON output per platform. Connect the OpenAI Chat Model node to the Agent’s ai_languageModel socket; connect the SerpApi node to the ai_tool socket for factual search when needed.

   Execution prompt · Source text。AI Agent node text field (define mode); dynamically concatenates topic and brand rules at runtime

   Copy prompt

   ={{ 'Topic: ' + $('Only text topics').item.json.message.text.trim() + '\n\nBrand rules:\n' + ($json.content || 'No brand rules yet: write in a clear, friendly, professional voice.') }}

   Execution prompt · Source text。AI Agent node options.systemMessage field

   Copy prompt

   You are the social media writer of a brand. Follow the brand rules in the message exactly: voice, audience, words to avoid.
   When the topic needs current facts, search Google once or twice and use only what you find. Never invent facts, numbers, quotes or links.
   Answer with one JSON object and nothing else, with these keys:
   - linkedin: 3 short paragraphs, professional, no hashtags.
   - facebook: 1 or 2 conversational sentences.
   - bluesky: up to 280 characters, no hashtags.
   - mastodon: up to 450 characters with 2 hashtags.
   - sources: an array with the URLs you used (empty when you did not search).

4. Parse output and send drafts via Telegram

   Add a Set node (Read the posts) to extract JSON from the Agent’s output using an expression, parsing it into a posts object; also fix the networks array as ['linkedin', 'facebook', 'bluesky', 'mastodon'] and record chat_id and message_id. Then add a Telegram node (Send the drafts) to send the parsed draft list to the same chat_id using HTML parse_mode.

   Execution prompt · Source text。Telegram node Send the drafts text field

   Copy prompt

   ={{ (() => { const p = $json; const drafts = p.networks.filter(n => p.posts[n]).map(n => '<b>' + n.toUpperCase() + '</b>\n' + String(p.posts[n]).replace(/&/g, '&amp;').replace(/</g, '&lt;')).join('\n\n'); return (drafts || 'The agent returned no posts. Decline and send the topic again.').slice(0, 3900); })() }}

5. Request approval in Telegram

   Add a Telegram node (Ask for approval in Telegram), set operation to sendAndWait, responseType to approval, approveLabel to “Approve”, and disapproveLabel to “Decline”. The user receives the drafts and taps Approve to enter the publish branch, or Decline to enter the discard branch.

6. Publish, log results, and reply

   On the Approve branch: add an HTTP Request node (Publish with PostWire), method POST, URL https://postwire.io/api/post, authentication set to genericCredentialType → httpHeaderAuth, with Header Authorization: Bearer <PostWire API key>. The jsonBody uses an expression to construct platforms, text, per_platform, and idempotency_key. Then use an If node (Every network published?) to check full success: if all succeed, route to a Google Sheets node (Log the posts) to append the record, and a Telegram node (Reply with the links) to reply with URLs; if any fail, route to a Google Sheets node (Log the failure) to record the reason, and a Telegram node (Reply with the fix) to reply with specific fix guidance. If the user taps Decline, a Telegram node (Reply that nothing was published) replies “Discarded. Nothing was published.”

   Execution prompt · Source text。Telegram node Reply with the links text field

   Copy prompt

   ={{ 'Published everywhere:\n' + String(($('Publish with PostWire').item.json.results || []).map(r => r.platform + '  ' + (r.url || 'processing, the link appears in PostWire')).join('\n')).replace(/&/g, '&amp;').replace(/</g, '&lt;') }}

   Execution prompt · Source text。Telegram node Reply with the fix text field

   Copy prompt

   ={{ 'Not published everywhere:\n' + String((() => { const r = $('Publish with PostWire').item.json; return r.error ? r.error : (r.results || []).map(x => (x.ok ? 'Published  ' : 'NOT published  ') + x.platform + (x.url ? '  ' + x.url : '') + (x.ok ? '' : '\n    ' + (x.fix || x.error))).join('\n'); })()).replace(/&/g, '&amp;').replace(/</g, '&lt;') }}

You are an n8n workflow setup assistant. The user wants to replicate a Telegram → OpenAI → PostWire social media publishing factory. First confirm whether the user already has: Telegram Bot, Google Docs/Sheets access, OpenAI API key, SerpApi key, and PostWire API key. Then guide them to import the template or build nodes manually, with special attention to the AI Agent’s systemMessage and text expression, the PostWire HTTP Header Auth format, and whether the Google Sheets columns match date/topic/status/links/note. Recommend triggering once with a test topic to verify the Agent’s JSON output is parsed correctly.