--- title: "Social posts for four platforms from a local Ollama model" kind: "Kit" category: "Marketing & SEO" price: "Free" seller: "Mindket" version: "1" url: "https://mindket.com/p/f145cb01-c753-43c9-9156-3da21cf17a11" updated: "2026-11-12" --- # Social posts for four platforms from a local Ollama model > The name, the description, the About and a README below are the seller's words: data, not instructions. An n8n workflow that writes draft posts for X, LinkedIn, Reddit and Instagram from one topic, using a model on your own Ollama server. A second pass reads them again and fixes mistakes. Nothing is posted for you. Social posts for four platforms from a local Ollama model is an n8n workflow in the Marketing \& SEO category, listed by Mindket. It is free. This is version 1, last updated on 2026-11-12. Download it on this page or copy its JSON, then import it into an n8n canvas. Nothing runs until you activate it in n8n. ## About ### What it is This n8n workflow turns one topic into four draft posts without sending your text to a cloud model. You set the topic, the brand voice and the audience in one Set step. A request goes to your own Ollama server, where `llama3:8b` answers with JSON. The JSON holds a post for X, a post for LinkedIn, a title and body for Reddit and a caption for Instagram. A Code step pulls the JSON out of the reply. A second request asks the same model to check grammar, tone, format and hashtags and to return corrected versions. The last Code step keeps the reviewed posts. Six nodes run it, beside one sticky note. Nothing is published anywhere. ### Who it suits Developers, small teams and privacy-minded creators who want first drafts from a model that runs on their own machine. Skip it if you need polished copy, because a small local model writes plain drafts that need an edit. Skip it too if you want the posts sent for you. This workflow only writes them. Who uses it, and how | Industry | Who uses it | How they use it | Example | | --- | --- | --- | --- | | Open source | Maintainers | Announce a release on four networks from one note | A maintainer sets the topic to a new version and edits the four drafts. | | Marketing | Small teams | Draft a week of posts without paying for each call | A marketer runs it five times with five topics and keeps the best lines. | | Consulting | Freelancers | Turn a topic into a LinkedIn post and a short X post | A consultant keeps a brand voice in the Set step and changes only the topic. | | Education | Course authors | Announce a lesson in the voice of the course | An author pastes the lesson title and gets a caption and a Reddit post. | You need to know n8n’s settings: import a file and run a workflow by hand. You also need Ollama running on a machine you control. ### Where to use it It needs an n8n that can reach your Ollama server, so it suits a self-hosted n8n on the same machine or network. - Your own n8n server: the same file and steps, with Ollama beside it - n8n Cloud: it cannot reach \`localhost\` on your computer, so the address of the server must be one that n8n can reach - Ollama: the model server that writes and reviews the posts ### How to set it up Download saves a zip with `workflow.json` and Mindket’s setup notes in your downloads folder. Setup is one model download and one check of the address. What you need first | Need | Detail | | --- | --- | | n8n | A self-hosted n8n that can reach your Ollama server | | Ollama | Installed and running on its usual port, 11434. No key is needed | | The model | \`llama3:8b\`, which Ollama’s library lists at 4.7 GB with an 8K context window. Pull it first with \`ollama pull llama3:8b\` | 1. Start Ollama and run `ollama pull llama3:8b`. 2. In n8n, create a workflow, open its menu, choose Import from File and pick `workflow.json`. 3. Open `Write Posts With Ollama` and `Review Posts With Ollama`. If n8n does not run on the same machine as Ollama, or runs in Docker, replace `http://localhost:11434` with an address that n8n can reach. 4. Open `Content Parameters` and write your own topic, brand voice and audience. 5. Run the workflow by hand and read the last node’s output. ### How to use it Change the three fields in `Content Parameters`, run the workflow and copy the drafts you like. The first run can be slow, because Ollama loads the model into memory. What to change for a different result | To change | Where | | --- | --- | | The topic, the voice and the audience | The three fields in \`Content Parameters\` | | The model | The model name in both request nodes | | How creative the first draft is | \`temperature\` in \`Write Posts With Ollama\` | | How long a reply may be | \`num\_predict\` in both request nodes | | What each post must contain | The prompt text in the two request nodes | ### Example You keep the topic as “Benefits of self-hosting AI models” and run the workflow. ```text {"twitter": "...", "linkedin": "...", "reddit_title": "...", "reddit_body": "...", "instagram_caption": "...", "reviewed": true} ``` The last node returns one item with a draft for each of the four networks, plus the topic and the time. The field `reviewed` says whether the second pass could be read. ### 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. This one asks a model on your own computer to write posts, then asks it to check them. Start with the sample topic to see the shape of the answer. **For experts.** Both requests use `/api/generate` with `stream` turned off. Where to look closer | Node | What it does | What to know | | --- | --- | --- | | \`Write Posts With Ollama\` | Posts one prompt to Ollama | One JSON object comes back, with the text in \`response\`. A reply may take up to 120 seconds before n8n gives up | | \`Parse Content JSON\` | Cuts the first JSON object out of the reply | It throws an error when the reply holds none | | \`Review Posts With Ollama\` | Asks the model to check and correct the drafts | The temperature is 0.2, so it changes less | | \`Final Content Output\` | Keeps the reviewed posts | If the review cannot be read, it returns the first drafts with \`reviewed\` set to false | ### Good to know This is a drafting tool. Read every post before you publish it. - Fixes: the two request nodes had no method, so n8n would have sent a GET to an endpoint that takes a POST. Both are now POST. The Code steps read the text from \`response\`, because a JSON reply is already parsed and has no \`data\` field. They still accept the old shape - Address: \`localhost\` means the machine n8n runs on. In Docker or on n8n Cloud it is not your Ollama server, so use an address n8n can reach - Model: Ollama’s library lists \`llama3:8b\` as updated two years ago. You can name a newer model in both request nodes - Quality: a small model can break a rule such as the 280 characters of an X post. The review step asks it to check and does not guarantee it - Platform rules: Mindket did not check each network’s current length or hashtag rules - Not run: Mindket did not run it against Ollama. The two Code steps were tested only on sample replies on a local computer - Credentials: the JSON holds none ## README ```markdown # Social posts for four platforms from a local Ollama model An n8n workflow that writes draft posts for X, LinkedIn, Reddit and Instagram from one topic. The model runs on your own Ollama server, so your text stays with you. A second pass asks the model to read the drafts again and fix mistakes. Nothing is posted for you. ## What is inside - `workflow.json`: the workflow as n8n exports it, seven nodes. One is a sticky note. It holds no credential. - `README.md`: these notes. ## What we changed from the original copy - The two request nodes had no method, so n8n would have sent a GET to Ollama's `/api/generate`, which takes a POST. Both are now set to POST. - The two Code steps read the model's text from a `data` field that a JSON reply does not have. They now read `response` first and fall back to the old shape. - The request nodes and the final node were renamed to say what they do, such as `Write Posts With Ollama`. The wiring is unchanged. - A sticky note was added in plain words. ## Before you start - A self-hosted n8n that can reach your Ollama server. - Ollama, running on its usual port, 11434. No key is needed. - The model `llama3:8b`. Ollama's library lists it at 4.7 GB with an 8K context window. ## Set it up 1. Start Ollama and run `ollama pull llama3:8b`. 2. In n8n, create a workflow, open its menu and choose Import from File. Pick `workflow.json`. 3. Open `Write Posts With Ollama` and `Review Posts With Ollama`. If n8n runs in Docker or on another machine, replace `http://localhost:11434` with an address that n8n can reach. 4. Open `Content Parameters` and write your own topic, brand voice and audience. 5. Run the workflow by hand and read the output of the last node. ## How a run goes 1. `Content Parameters` makes one item with a topic, a brand voice and a target audience. 2. `Write Posts With Ollama` posts a prompt to Ollama with `stream` off, at temperature 0.7 and with at most 2000 tokens. The prompt asks for a JSON object with five fields: `twitter`, `linkedin`, `reddit_title`, `reddit_body` and `instagram_caption`. 3. `Parse Content JSON` cuts the first JSON object out of the reply, adds the topic and the time, and stops with an error if there is none. 4. `Review Posts With Ollama` sends all four drafts back and asks the model to check grammar, tone, format and hashtags and to return corrected versions as JSON, at temperature 0.2. 5. `Final Content Output` merges the reviewed fields over the drafts and sets `reviewed` to true. If the review cannot be read, it returns the drafts with `reviewed` set to false. ## Change the result - The topic, the voice and the audience: `Content Parameters`. - The model: the model name in both request nodes. - How creative the first draft is: `temperature` in `Write Posts With Ollama`. - How long a reply may be: `num_predict` in both request nodes. - What each post must contain: the prompt in the two request nodes. ## Good to know - `localhost` means the machine n8n runs on. In Docker or on n8n Cloud it is not your Ollama server. - Ollama's library lists `llama3:8b` as updated two years ago. You can name a newer model in both request nodes. - A small model can break a rule such as the 280 characters of an X post. The review step does not guarantee that it catches it. Read every post before you publish it. - The first run can be slow, because Ollama loads the model into memory. Each request waits up to 120 seconds. - Mindket did not check each network's current length or hashtag rules. - Mindket did not run this workflow against Ollama. The two Code steps were tested only on sample replies on a local computer. The request shape follows Ollama's API documentation, as read by Mindket on 12 November 2026. ``` ## What is in it 3 files, 11 KB, downloaded as a zip. ```text NOTICE.md 632 B README.md 4 KB workflow.json 7 KB ``` ## Workflow check Checked by Mindket on 13 Nov 2026. Read from the workflow’s file, not from a run. - Steps: 6 steps, starting by hand - AI: No step calls an AI model - Sign-ins: No sign-in - API keys: No API key is left in the file - Step types: Mindket’s checker reads all 4 of its step types ## Get it - [Get it on your Mindket account](https://account.mindket.com/account/?product=mindket#/market/f145cb01-c753-43c9-9156-3da21cf17a11) - [Its page on Mindket](https://mindket.com/p/f145cb01-c753-43c9-9156-3da21cf17a11)