How To Post To Social Media With AI
If you want to post to social media with AI, the shortest path is to connect your accounts, run an MCP server, and let the agent call a real posting tool. That is what sfeed does for Facebook Pages and Instagram.
The short answer
curl -fsSL https://sfeed.dev/install.sh | sh
sfeed auth facebook
sfeed mcp After that, your agent can inspect the connected Pages, choose the right one, publish immediately, or schedule posts for later.
You can keep your posts, media, prompts, and schedules however you want on your own machine. Organize them in folders, Markdown files, scripts, or whatever workflow you already use, then tell Claude, Codex, OpenCode, or another MCP client to use sfeed to publish them according to your rules.
Local files are the source of truth
content/
rules.md
queue/
2026-04-10-product-post.md
2026-04-11-instagram-post.md
media/
launch-card.jpg
demo.mp4 This is the core idea: your files stay local, your agent reads them, and sfeed is the publishing layer. You do not need to move your content into a hosted dashboard first.
See Example filesystem setup for a fuller version.
What sfeed handles
- Browser auth for Facebook Pages and linked Instagram accounts
- Immediate posting from the terminal or through MCP
- Hosted scheduling when the post should run later
- Browser previews and a hosted queue UI for scheduled posts
- Queue actions like reschedule, duplicate, and cancel
Typical AI-agent flow
- You run
sfeed auth facebookonce - You run
sfeed mcp - Your agent reads your local files and rules
- Your agent checks status and connected Pages
- Your agent drafts the post
- Your agent publishes it now or schedules it for later
Example prompts
Post this to Facebook: "Shipping v0.1.3 today."
Schedule this Instagram post for tomorrow at 9am with ./launch-card.jpg
Show me my scheduled posts and open the preview for the newest one For a more realistic flow, tell the agent to read your local content folder and follow your rules file before it posts anything.
Why use sfeed instead of building it yourself
Most AI agents can write text, but they still need a real integration to connect accounts, upload media, schedule posts, and keep a queue. sfeed gives the agent that integration through the CLI and MCP, while you keep control of how the source material is organized on your own machine.