How I Run My Social Media With Claude Code and Blotato
I built an AI marketing officer in Claude Code that drafts a week of posts in my voice, runs quality gates, and publishes through Blotato.
I use Claude Code for social media the way most people wish their content team worked. I type one sentence, and it researches the topic, drafts a week of posts in my voice, generates the visuals, then waits for me to review before anything goes out.
This is the system behind the reach I get across my channels. It is the definition of batch producing content and using AI to actually scale yourself, instead of staring at a blank calendar every Monday.
The payoff is simple. One person can put out the volume that used to take a small team, without the content turning into lazy AI slop.
In this post I walk through the exact build from my full course: installing Claude Code, creating skills, wiring up quality gates, training brand voice, and using parallel sub-agents to publish everywhere at once. Most guides on this topic stop at “draft your content.” The hard part is everything after the draft, and that is where this build earns its keep.
How to Use Claude Code for Social Media (Video Guide)
If you would rather watch the full walkthrough, this is the build-along version where I create the whole system on camera. The written guide below covers the same build with extra detail on the gotchas and the parts that broke live.
Why Most AI Content Workflows Stall at the Draft
Most AI content advice gives you a prompt and a pat on the back. You paste it into a chat, get a post, and copy it somewhere by hand. That is fine for one post, and if you are just starting out, a handful of copy-paste Claude prompts will get you surprisingly far. It falls apart the second you want a week of content across several platforms.
The blank slate is not the real problem. The real problem is the work that comes after a draft exists: checking it against your brand voice, catching the formatting that breaks on each platform, and getting it published without 40 copy-paste steps.
I have spent over 500 hours in Claude Code building my product, and the same engine that writes code can run that whole back half for you. Built right, it researches, drafts, checks quality, and publishes. You stay in the loop where it matters and skip the busywork where it doesn’t. If you want the deeper engineering view of that same back half, I broke it down in my AI content pipeline in Claude Code guide.
My Claude Code Social Media Stack
The build is simpler than it sounds. Here is what each piece does.
- Claude Code: the agent that runs everything. I run it inside the Visual Studio Code extension and use Opus 4.6 for the technical work.
- Skills: custom slash commands you build once and reuse forever. A skill lives in your project as a
.claude/skills/directory with aSKILL.mdfile of instructions. I build two:/postfor a single post, and/plan-weekfor a full week. - Hooks: small scripts that run automatically before an action. I use one as a quality gate that blocks any post that breaks the rules before it can publish.
- Blotato: the app I built so I could make a ton of content without going insane. Claude calls the Blotato API to generate visuals and publish to my social accounts. I’m involved with Blotato as a creator and tester, so take this with whatever grain of salt feels right.
The piece that breaks most DIY versions of this build is the publishing step. Wiring up each platform API by hand is where people give up, which is why every workflow here hands the publish step to Blotato. If you want to run the same stack on your own accounts, start a free 7-day Blotato trial, connect a platform, and run your first AI-drafted post end to end. The trial covers the API access, the visual generation, and the publishing endpoints this whole build depends on.
Claude Code vs No-Code Automation: What Actually Fits
Before the build, here is how I think about Claude Code against the no-code routes people usually compare it to. None of these is strictly better. They fit different jobs.
| Approach | Best at | Weak spot |
|---|---|---|
| Claude Code | Drafting in your voice, quality checks, flexible logic | Orchestration, scheduling, pre-built integrations |
| n8n | Orchestration, scheduled triggers, deployment | Steeper setup, you build the workflow first |
| Make.com | Visual scenarios, cheaper than most | Per-operation cost adds up at posting volume |
| Web chat alone | Quick one-off drafts | No publishing, no memory, no quality gate |
I see Claude Code and n8n as complementary tools, not rivals. There is a lot in the orchestration layer that n8n does nicely that Claude Code is pretty weak at. You can even connect the two over MCP so Claude Code fixes your n8n automation when it breaks. If you are weighing the no-code side, I would check out Make.com over Zapier because it is significantly cheaper, and you can dig deeper in my n8n vs Make vs MCP breakdown.
How to Build the System in Claude Code (Step-by-Step)
Here is the build in the order I do it on camera. Each step stacks on the last.
Step 1: Install Claude Code and Set Permissions
Sign up at claude.ai, install Visual Studio Code, then add the official Claude Code extension. Search “Claude” in the Extensions panel, and the first result with over 4 million downloads is the one you want. Log in with /login, then create a project folder. I named mine “AI Social Media Manager.”
The first thing I do is set permissions so Claude can run the safe, non-destructive commands without asking every single time. I spend about 90% of my time in plan mode so Claude doesn’t accidentally change anything, then switch to edit automatically once I am happy with the plan.
Step 2: Create the /post Skill
I tell Claude Code to build a skill named post: an AI social media manager that takes a topic, a draft, or a URL, parses the intent, pulls content through the Blotato source API, creates a visual, and publishes to my socials. I end the prompt with one line that changes everything: “ask me clarifying questions one at a time until you’re 95% confident you can complete the task successfully.”
That single instruction catches the blind spots. No matter how good you think your prompt is, there are always edge cases you didn’t think about. Claude then generates the SKILL.md with the full instruction set. The exact prompts I run to drive a full content session live in my Claude Code prompts breakdown.
Step 3: Add Your Blotato API Key
In Blotato, go to Settings, then API, and generate a key. That is a paid feature. Paste it into a settings.local.json file as an environment variable so it never ends up in your shared config. Now the skill can actually reach your accounts.
Step 4: Test the Post Skill
I run /post with a topic and a YouTube URL. Claude pulls the transcript, then drafts three Twitter options I can choose from. It can analyze a two-hour YouTube video, which I honestly didn’t think would work, and it worked insanely well. If you give it a topic with no URL, it runs a research step first, then drafts.
Step 5: Train Your Brand Voice
A generic AI post sounds like a generic AI post. I paste in a humanizing prompt, then add voice samples per platform. If you have no samples yet, you can have Claude look up a creator you admire as stylistic inspiration. I used Alex Hormozi for Twitter and Justin Welsh for LinkedIn. Claude fetches their public posts and updates the brand-voice section of the skill.
Step 6: Add a Quality Gate With a Hook
This is my favorite part. I create a Claude Code hook, a script that runs before publishing, and I call it a quality gate. As soon as you go to publish something, you have to cross this gate. It checks for em dashes, platform character limits, a required image or video, and banned words. The hook file lives in a hooks/ directory and runs before any publish command. I test it with an oversized tweet, and the gate blocks it with a clear reason.
Step 7: Parallelize With Sub-Agents
Once one platform works, I update the skill so it can fan out. Typing “all” spawns a sub-agent per platform, each one adapting the post to that platform’s format and publishing at the same time. Every result gets written to a posts log so I have a record of what went where. In the demo I publish to Twitter and LinkedIn from a single visual.
Step 8: Build the /plan-week Skill and Schedule
The second skill plans a whole week. It takes a topic or URL, researches it, then generates 10 angles and 10 full drafts across platforms, plus a few infographics. It writes a content-plan file I can read and edit, waits for my approval, and only then schedules each post into the next free slot. You set those slots up in the Blotato calendar so the content fills your weekly schedule automatically.
I save the CLAUDE.md file for last. It codifies everything you built: where the API key lives, the brand-voice file, your conventions, the endpoints, the scheduling rules, and the quality gates. I am kind of religious about my CLAUDE.md file because it keeps every future session on the rails.
Pro Tips From Building This Live
Clear context between features. I clear the context every single feature or bug. I get much better results when Claude starts with a clean slate instead of dragging an hour of unrelated history into a new task.
Keep a human in the loop, especially early. Most people should not start out with full automation of content. If you don’t know how to make good content yet, you are not going to learn it by setting and forgetting. The review step is the point, not the obstacle.
Match your posting cadence to the platform. On Twitter you can post 15 times a day and not get punished. On LinkedIn you generally don’t want more than once per day, and three posts a day on Instagram is already pushing it.
What This Workflow Can’t Do (Yet)
It is not magic, and a few things bit me on camera. The Blotato API is a paid feature, so you need credits, and I actually ran out mid-demo. Scheduling has sharp edges too: on one /plan-week run, the skill posted everything immediately instead of scheduling it into the future, which is a prompt bug I had to flag and fix before relying on it.
Claude Code is also weak at orchestration, deployment, and pre-built integrations. If you need scheduled triggers and connectors that just exist, that is where pairing it with n8n earns its place. And the quality gate only catches what you tell it to catch, so a sloppy rule list still ships sloppy posts.
Results You Can Expect
Once this is running, a single sentence turns into a reviewed, on-brand week of content across your platforms. The /plan-week skill alone saves me at least 5 hours every week, and that is before counting the posts I would have skipped entirely on a busy day.
The honest expectation is a force multiplier, not a hands-off machine. You still approve the plan. You just stop doing the copy-paste, the reformatting, and the platform-by-platform publishing by hand.
Sabrina’s Final Take
If you already make decent content and you are drowning in the production side, this build is worth the weekend it takes to set up. Start with the /post skill, get one platform publishing cleanly, then add the quality gate and the second skill once you trust it. Claude Code is pretty addictive once you see what you can build, so resist the urge to automate everything on day one. Every Blotato plan covers the API access and publishing this stack runs on, so the 7-day trial is enough to test the whole thing end to end before you commit.
Claude Code for Social Media FAQs
Can Claude Code post to social media directly?
Not on its own. Claude Code drafts and decides, but it needs a publishing layer to actually post. In this build it calls the Blotato API to push finished posts to your connected accounts, which is how it reaches Twitter, LinkedIn, and the rest from one command.
Do I need to know how to code to build this?
No. The whole system is built by describing what you want in plain English and letting Claude Code write the files. Skills, hooks, and the config are all generated from prompts. The only technical step is pasting an API key into a settings file. Sandy Lee had never written a line of code six months before she was running a $48K month on this exact stack.
What does Claude Code cost to run this?
I am on the 200 dollar a month Claude plan because I get so much value from it, and I recommend at least the 100 dollar tier for serious building. You also need Blotato credits for the visual generation and publishing, which is a separate paid piece.
Do I have to connect every platform at once?
No, and I would not. Start with the one or two accounts you actually post on, get the /post skill publishing there cleanly, then connect more as you trust the gate. Each new platform is just another connected account the sub-agents can target. There is nothing to rebuild when you add one.
Should I fully automate my content with this?
I would not, at least not at the start. Keep a human approval step in the plan-week flow. The system is built to multiply your output while you stay in control of quality, not to remove you from the process entirely.