How to Automate Social Media With Claude Code (2026)

You can automate social media with Claude Code by installing the open-source simplified-cli plugin, setting your Simplified API key, then telling Claude what you want in plain English. Type “generate a sunset image and post it to Instagram,” and the agent runs the right commands, waits for the image to render, and queues the post. No dashboard, no tab-switching, no manual export.

That’s the short version. Here’s the part that matters: this isn’t a chatbot that writes captions and stops. Claude Code is Anthropic’s agentic coding CLI, and with the right plugin it can actually take action across 10 social platforms. You describe the outcome, it figures out the steps.

This guide walks through the full setup and four real workflows you can copy today. We’ll show the exact prompts you type to Claude and the underlying simplified commands the agent runs, so you can see what’s happening under the hood and reproduce it on your own accounts.

Key Takeaways

• To automate social media with Claude Code, install the simplified-cli plugin and set SIMPLIFIED_API_KEY. Then describe tasks in plain English and the agent runs the commands.
• The CLI ships with a Claude Code skill, so the agent already knows which commands to run for posting, scheduling, image generation, and analytics.
• It connects to 10 platforms: Instagram, Facebook, LinkedIn, TikTok, YouTube, Pinterest, Threads, Google Business, Bluesky, and TikTok Business.
• Real workflows take one prompt: generate-and-post an image, schedule a week of posts, pull last month’s analytics, or turn a blog into social posts.
• Everything outputs JSON, so the agent reads results and chains the next step without you copying data between tools.

Why automate social media with an AI agent instead of a dashboard

Dashboards make you do the work. You open the tab, pick the account, write the caption, upload the media, set the time, and hit schedule. Repeat for every post, every platform, every week. The tool helps, but you’re still the one clicking.

An AI agent flips that. You state the goal once, and it executes the steps. The difference is who holds the workflow in their head: you, or the agent.

Here’s a concrete before-and-after. Posting an image to Instagram from a dashboard means generating the image somewhere, downloading it, uploading it, writing the caption, and scheduling. Five steps, maybe three minutes, across two tools. With Claude Code and simplified-cli, you type one sentence and the agent does all five.

A few reasons this approach wins for small teams:

• No context-switching. You stay in one terminal. The agent talks to Simplified through the CLI, so design, writing, and scheduling happen in the same place.
• It chains tasks. Because every command returns JSON, the agent reads the result of one step and feeds it into the next. Generate an image, grab the URL it returns, then post with that URL—automatically.
• It handles the boring parts. Pulling last month’s numbers, reformatting a blog into five posts, queueing a week of content. These are repetitive and rule-based, exactly what an agent is good at.
• You stay in control. The agent proposes the commands and runs them, but you approve the actions and review the output. It’s automation with a human in the loop, not a black box.

If you want the bigger picture on where this fits, our social media automation guide covers the strategy side. This article is the hands-on how-to.

What you need before you start

Three things. None of them take long to set up.

1. Claude Code installed. Claude Code is Anthropic’s agentic CLI for the terminal. If you’ve used it for coding, this is the same tool—you’re just pointing it at social media instead of a codebase. It supports plugins and MCP (Model Context Protocol), which is how external tools like simplified-cli connect to it. Check Anthropic’s documentation if you haven’t installed it yet.

2. A Simplified account. You’ll need an active account to connect your social profiles and generate content. The free plan lets you connect 3 social accounts and test the workflow before you commit to anything. Connect your Instagram, LinkedIn, or whichever platforms you post to.

3. A Simplified API key. This is what authorizes the CLI to act on your account. Go to Simplified.com, open Settings, find API Keys, and generate one. Copy it somewhere safe—you’ll set it as an environment variable in a moment.

That’s the whole list. You don’t need to write code, understand MCP internals, or read API documentation. The plugin and its bundled skill handle the technical details.

Install the simplified-cli plugin and set your API key

There are two ways to install simplified-cli. Pick whichever fits how you work.

Option A: Install as a Claude Code plugin (recommended)

Inside Claude Code, add the marketplace and install the plugin:

/plugin marketplace add celeryhq/simplified-cli

This registers the plugin with Claude Code. The big advantage here: the plugin ships with a Claude Code skill, a bundled instruction set that teaches the agent exactly how to use every command. You don’t have to explain how posting or analytics work. The agent already knows.

Option B: Install via npm

If you’d rather have the binary available system-wide, install it globally with npm:

npm install -g simplified-cli

This gives you a simplified command you can run directly in any terminal, and Claude Code can call it too. Use this if you also want to script tasks outside of Claude.

Set your API key

Either way, the CLI needs your API key to authenticate. Export it as an environment variable:

export SIMPLIFIED_API_KEY=your_key_here

That’s it. To confirm everything connected, ask Claude to list your accounts:

You type: “List my connected social accounts.”

Claude runs:

simplified accounts:list

The command returns JSON with each connected account and its ID. You’ll use those account IDs to target specific platforms. Keep them handy—or just let the agent look them up each time, since it can re-run accounts:list whenever it needs an ID.

Want to understand how this compares to calling the API directly? Our guide on posting to social media via API covers the raw approach. The CLI wraps all of that so you don’t have to.

Real workflows: prompts you type and commands Claude runs

Here’s where it gets practical. Each workflow below shows the plain-English prompt you give Claude and the simplified commands the agent runs to make it happen. These are reproducible—set up the plugin, then try them on your own accounts.

Workflow 1: Generate an image and post it to Instagram

This is the headline use case: one prompt, image to published queue.

You type: “Generate a realistic sunset image over a city skyline, then post it to my Instagram with the caption ‘Golden hour in the city.'”

The agent breaks this into two steps. First it generates the image. Because image generation runs asynchronously, it returns a task_id, so the agent uses --wait to hold until the image is ready:

simplified ai-image:generate --model flux.flux-realism --prompt "realistic sunset over a city skyline, golden hour" --wait

When that finishes, the command returns the image URL in its JSON output. The agent grabs that URL and creates the post, queuing it to your Instagram account:

simplified posts:create -c "Golden hour in the city." -a "<instagram_account_id>" --action add_to_queue --media "<generated_image_url>"

Two commands, one sentence from you. The agent handled the handoff between image generation and posting on its own. If you want a different look, you change the prompt—”minimalist flat illustration” or “moody black-and-white photo”—and the agent adjusts the --prompt value.

Workflow 2: Schedule a week of posts

Planning ahead is where automation earns its keep. Instead of scheduling seven posts by hand, describe the week and let the agent queue them.

You type: “Draft 5 LinkedIn posts about remote team productivity, one per weekday, and add them all to my queue.”

The agent writes five captions (this is where Claude’s writing ability does the heavy lifting), then runs posts:create once per post, adding each to the queue:

simplified posts:create -c "Remote work isn't about location. It's about trust..." -a "<linkedin_account_id>" --action add_to_queue
simplified posts:create -c "Async updates beat status meetings for one reason..." -a "<linkedin_account_id>" --action add_to_queue

…and so on for all five. Each call returns JSON confirming the post was queued, so the agent can report back exactly what it scheduled.

Need to change one? Ask Claude to list your scheduled posts and update the one you want:

You type: “Show me my queued LinkedIn posts and reword Wednesday’s to be more casual.”

Claude runs:

simplified posts:list -a "<linkedin_account_id>"
simplified posts:update <post_id> -c "New casual caption..."

You can delete with posts:delete <post_id> the same way. For more on crafting the captions themselves, our AI social media post generator guide goes deeper on writing posts that land.

Workflow 3: Pull last month’s analytics

Reporting is the chore nobody wants. The agent makes it a single question.

You type: “Pull my Instagram analytics for May and tell me which posts performed best.”

The agent first grabs the aggregated numbers for the date range, then pulls per-post data to find the winners:

simplified analytics:aggregated -a "<instagram_account_id>" --from 2026-05-01 --to 2026-05-31
simplified analytics:posts -a "<instagram_account_id>" --from 2026-05-01 --to 2026-05-31

Both return JSON. Because the agent can read that output directly, it doesn’t just dump numbers at you—it summarizes reach, engagement, and your top posts in plain language. Want to know who you’re reaching? Add audience data:

simplified analytics:audience -a "<instagram_account_id>"

Now you can ask follow-ups in the same conversation: “Which of these should I repost?” or “What time were my best posts published?” The agent already has the data loaded, so it answers without you re-pulling anything.

Workflow 4: Repurpose a blog post into social posts

You wrote a blog post. Now you need it on five platforms in five different formats. This is a perfect agent task because it combines writing, image generation, and posting.

You type: “Take my latest blog post on email subject lines, turn it into a LinkedIn post, an Instagram caption with a matching image, and a Bluesky post. Queue all three.”

The agent reads the source content (paste it or point Claude at the file), writes platform-specific versions, generates an image for the visual platforms, then queues each post:

simplified ai-image:generate --model flux.flux-realism --prompt "clean graphic about writing better email subject lines" --wait
simplified posts:create -c "<LinkedIn version>" -a "<linkedin_account_id>" --action add_to_queue
simplified posts:create -c "<Instagram caption>" -a "<instagram_account_id>" --action add_to_queue --media "<image_url>"
simplified posts:create -c "<Bluesky version>" -a "<bluesky_account_id>" --action add_to_queue

One blog post, three platforms, one prompt. The agent adapted the tone for each channel—punchy for Bluesky, professional for LinkedIn, visual-first for Instagram—and handled the image generation in between. For the strategy behind repurposing, see our AI agents for social media guide.

Beyond these four, the CLI also handles video and image editing. You can run simplified video:script-to-video --wait to turn a script into a video, or simplified image:remove-background --wait to clean up a product photo before posting. Same pattern every time: async tasks return a task_id, and --wait holds until the result is ready.

Tips and guardrails for using Claude Code safely

Automation is powerful, which means a sloppy prompt can publish something you didn’t intend. A few habits keep you in control.

Use the queue, not instant publish. The --action add_to_queue flag adds posts to your Simplified queue instead of pushing them live immediately. This gives you a review step. Check the queue in Simplified before anything goes out, especially when you’re getting started.

Confirm account IDs before bulk actions. When you schedule a week of posts, make sure the agent is targeting the right account. A quick “list my accounts first, then schedule” in your prompt avoids posting LinkedIn content to your Instagram by mistake.

Review captions before queueing. Claude writes good first drafts, but it’s still a draft. Ask to see the captions before they’re queued: “Show me the five posts before you add them.” Read them, tweak what’s off, then approve.

Keep your API key private. Your SIMPLIFIED_API_KEY controls your account. Don’t paste it into shared chats, commit it to a repo, or hardcode it into scripts. Set it as an environment variable and leave it there.

Always --wait on async tasks. Image generation, video creation, and background removal run asynchronously. If you don’t wait, the agent might try to post before the media exists. The bundled skill knows to use --wait, but it’s worth confirming when you write custom prompts.

Be honest about what this is. This setup automates the mechanical work—generating, writing drafts, scheduling, reporting. It doesn’t replace your judgment about what to say or when to say it. Treat the agent as a fast assistant, not an autopilot. You’re still the editor.

One more practical note: the agent is only as good as the prompt. “Post something” gives you something generic. “Write a LinkedIn post about our new feature, keep it under 100 words, end with a question” gives you something usable. Specific prompts, specific results.

Start automating your social media with Claude Code

Automating social media with Claude Code comes down to four moves. Install the simplified-cli plugin with /plugin marketplace add celeryhq/simplified-cli. Set your SIMPLIFIED_API_KEY. Confirm your accounts with accounts:list. Then describe what you want and let the agent run the commands.

The payoff is real for small teams. Instead of clicking through a dashboard for every post, you generate an image, schedule a week of content, pull a month of analytics, or repurpose a blog—each with a single plain-English prompt. The agent handles the JSON handoffs, the --wait timing, and the per-platform formatting. You handle the strategy and the final approval.

Is this for everyone? If you live in a visual dashboard and never touch a terminal, the standard Simplified app is the better fit. But if you’re comfortable in Claude Code and want your social workflow to run on plain English, the CLI turns “I should post this” into one sentence.

Frequently asked questions

How do I automate social media with Claude Code?

Install the simplified-cli plugin in Claude Code with /plugin marketplace add celeryhq/simplified-cli, then set your Simplified API key with export SIMPLIFIED_API_KEY=your_key. After that, type plain-English requests like “generate an image and post it to Instagram,” and the agent runs the right simplified commands for you, including image generation, scheduling, and analytics.

Do I need to know how to code to use the simplified-cli plugin?

No. The plugin ships with a Claude Code skill that teaches the agent every command, so you describe tasks in natural language and Claude handles the commands. You set the API key once as an environment variable, which is a single copy-paste. You never write code or read API documentation yourself.

What’s the difference between Claude Code plugins and MCP for social media?

Claude Code plugins (like simplified-cli) bundle commands and a skill that the agent loads directly, while MCP (Model Context Protocol) is the broader standard for connecting external tools to Claude. The simplified-cli plugin is the simplest path for social media because it installs in one line and the agent already knows how to use it.

Which social platforms does simplified-cli support?

The CLI connects to 10 platforms through Simplified: Instagram, Facebook, LinkedIn, TikTok, YouTube, Pinterest, Threads, Google Business, Bluesky, and TikTok Business. You target a specific platform using its account ID, which you get from the simplified accounts:list command.

Can Claude Code generate images and post them automatically?

Yes. Ask Claude to generate an image and post it, and it runs simplified ai-image:generate --model flux.flux-realism --prompt "..." --wait to create the image, then passes the returned URL into simplified posts:create ... --media "<url>" to queue the post. The --wait flag holds until the image is ready so nothing publishes before the media exists.