Launching today

pendpost

Launching today

Agent-operated social media with a human approval gate.

4 followers

Agent-operated social media with a human approval gate.

4 followers

Visit website

A free, open-source, local-first social planner where an AI agent drafts and schedules posts to Instagram, LinkedIn, YouTube, and X behind a human approval gate you control. MCP-native: drive it from Claude, Cursor, or the dashboard. MIT-license. npx pendpost.

Free

Launch tags:Social Media•Developer Tools•Artificial Intelligence

Launch Team

Fin Startups get Fin free for a year + 93% off Intercom

Promoted

Hunter

📌

Hi everyone, maker here. I built pendpost because I wanted agents to run my social media without handing them the keys to my accounts. Most "AI social" tools do one of two things. They post on your behalf with no real guardrail, or they are plain schedulers that assume a human wrote every post. Neither fits the way I work now, where an agent does the drafting and I still want to be responsible for what goes out. So pendpost puts a gate in the middle: - An agent (Claude, Cursor, anything that speaks MCP) drafts and schedules across Instagram, Facebook, LinkedIn, YouTube, and X. You drive it from your agent, or from a local dashboard that does everything the agent can. - Every post is fail-closed by default: nothing publishes until it is approved. - When you trust it, you opt into auto-approve and go more hands-off, all the way to fully automated. You turn it on, you can switch it off, and only you can do either. - It is local-first and MIT. It runs on your machine, your keys stay in your own .env, nothing phones home. - There are anti-ban circuit breakers and a caption brand-lint, so you are less likely to get flagged or ship robotic copy. If you would rather not keep a machine on, there is an optional cloud that keeps your lanes publishing 24/7 with the same gate enforced on the server. The local core stays free and complete either way. Two things worth saying. This launch post was itself drafted and scheduled through pendpost, and the project is young, so it has rough edges. Try the whole loop in mock mode with no credentials: npx pendpost I would love feedback, especially from anyone running agents against real accounts. What would it take for you to trust an agent to post for you?

Report

11d ago

How does the approval gate actually work in practice, does the agent pause and ping me for each post or batch them up by day so I'm not constantly approving one by one?

Report

2h ago

Hunter

@fatih1pmo
Posts land as drafts in a review queue and nothing publishes until you say so. You can approve it whenever suits you, approve a batch in one sitting or one at a time. You can prompt your agent to tell you what is open to approve or you watch it all from the desktop app where you have a real GUI to review, monitor what's queued, and see what already went out.

The approval-gate can be customized by you. You can keep everything manual, or pre-authorize auto-approval scoped to what you trust (for example, routine LinkedIn text posts flow through but every Reel still stops for you).

You can also manually add or edit posts in the app. Because the whole thing runs over MCP, your agent can draft and schedule straight into that same queue. You choose if you stay the approver or let your agent go fully autonomous.

Report

2h ago