X post writer with Telegram approval

What it is

This n8n workflow writes posts for X and waits for your approval before it publishes them. You message your Telegram bot with what the post should cover. An agent searches the web through Tavily, writes a draft of up to 280 characters with its source named, and sends it back to your chat. The file holds two complete flows for that approval step. In one you press Approve or Decline. In the other you type a reply, a classifier decides whether it means yes, and a request for changes goes to a revision agent, which sends a new draft for another round. Approved text is posted to X.

1. You message your Telegram bot
2. An agent searches the web and drafts the post
3. The draft returns to your chat for approval
4. Approved text is posted to X

Who it suits

People who post on X often, want a draft made from fresh sources, and want to keep the last word. Because the whole loop runs inside a chat, it works from a phone.

Skip it if you want posts on a schedule or a content calendar. This answers one message at a time and has no queue.

Where to use it

An n8n with a public HTTPS address, because Telegram sends your bot’s messages to it.

• n8n Cloud or your own server: import the file and publish it, so the bot’s messages reach the triggers
• Telegram: the chat with your own bot is where requests, drafts and replies happen
• X: approved posts go to the account you connect
• Search and models: Tavily finds the sources and OpenRouter runs the language models

Setup

Get it saves a zip with workflow.json and Mindket’s setup notes in your downloads folder. Have these ready before you import it.

1. In n8n, create a workflow, open its menu, choose Import from File and pick workflow.json.
2. Add a Telegram credential with your bot token and select it in Telegram Trigger, Submit Approval, Denial message, Telegram_Trigger and Request Feedback.
3. Add an HTTP Header Auth credential with the header Authorization and your Tavily key, and select it in Tavily and Tavily Search.
4. Add an OpenRouter credential and select it in GPT 4.1, GPT_4.1 and 2.0 Flash.
5. In the X developer portal, make an OAuth 2.0 app and give it the callback address n8n shows on the credential. Add an X OAuth2 credential with the Client ID and Secret, then select it in X and X Post.
6. Read the note on bots under Good to know, then choose which flow to keep.
7. Save, press Publish, and message your bot.

How to use it

Message the bot with what the post should cover. The agent searches, drafts, and sends the draft back to the same chat. How you answer depends on the flow.

Example

A message to the bot, and the typed-reply flow:

You: a post about this week’s Mars rover news
Bot: Good to go? <the draft, up to 280 characters, with its source named>
You: make it shorter
Bot: Good to go? <the shorter draft>
You: looks good

The last reply makes the classifier answer approved, and the latest draft is posted to X.

For newcomers and experts

If you are new. n8n draws a workflow as boxes joined by lines, and each box, a node, does one job. Notes on the canvas group this file into its two flows and label each stage, such as Initial Content, HITL, Decision Point, Post and Revision. Start with the yes or no flow, since it has fewer parts.

For experts. Models run through OpenRouter, and each role has its own node.

Good to know

A post is public, so the approval step is the only thing between a draft and your followers.

• One bot, two triggers: Telegram sends a bot’s messages to a single webhook address, so two triggers on the same bot compete. Give the second flow its own bot, or delete the flow you do not use
• Length: the prompt asks for 280 characters or fewer, and nothing in the workflow checks it
• Waiting: neither approval step sets a time limit, so an unanswered draft keeps waiting
• Replies: the answer opens a page served by your n8n, so the reviewer’s phone must be able to reach its address
• Costs: OpenRouter bills each draft and each classification, and Tavily bills each search
• Your data: your request and the drafts reach Tavily, OpenRouter and the model providers, and approved text reaches X