The social publishing layer for AI agents
One connection. Every account. Any agent.
Connect Postproxy once, and any MCP-aware agent can publish and schedule across 11 platforms. It can also reply to comments and DMs where the platform supports them. Works with Claude, ChatGPT, Cursor, Codex, Hermes, and OpenClaw. Your credentials stay with you.
Find your setup
Pick where your agent already lives. Each path is one connection, then plain-language publishing — the agent calls Postproxy while your credentials stay server-side.
Claude
One claude mcp add, or connect from the Claude directory in a click.
ChatGPT
Paste the connector URL in Settings → Apps and sign in.
Cursor
One block in .cursor/mcp.json, committed with the repo.
Codex
Paste the connector URL in Settings → Apps and sign in.
Hermes
Install the skill, drop in one API key, publish from any chat.
OpenClaw
Install the skill from ClawHub; every routed channel gets the tools.
MCP
The protocol behind it all — hosted endpoint or local npm package.
Three steps from agent to published
The same flow for every client. Connect once, approve access, then ask in plain language.
- 1
Point your client at the MCP endpoint
Register `https://mcp.postproxy.dev/mcp` in your agent — one command, one config block, or one click in a connector directory.
- 2
Authorize with your Postproxy account
Sign in over OAuth or drop in an API key. Platform credentials stay in Postproxy, so the agent works with tools while tokens stay server-side.
- 3
Ask for what you want, in plain language
"Schedule Thursday's launch post to X and LinkedIn." The agent calls the tools; Postproxy handles formatting, media, rate limits, retries, and reports back.
We own the hard parts
Your agent focuses on what to say. Postproxy handles delivery, feedback, and every platform's quirks behind one stable contract.
Automatic retries with back-off on every publish, so a transient platform failure recovers on its own instead of dropping your post.
Delivery status, incoming comments, DMs, and reviews come back as structured results — the agent knows what actually happened.
Per-platform limits are tracked on our side. Your agent calls once and Postproxy queues intelligently.
Pull post and follower stats back after publishing, normalized across networks, ready for the next decision.
When Instagram, LinkedIn, or TikTok shift their API, we absorb it — your agent keeps working while we adapt.
One OAuth sign-in or API key connects everything, and platform tokens stay server-side.
Built for agentic workflows.
We know your stack: MCP, skills, n8n, Make. We're built to fit right in and keep publishing smooth.
Give your agent everything it needs to publish and respond to social content from your project.
npx skills add postproxy/postproxy-skill 
MCP
Postproxy MCP server lets Claude, Cursor, and other MCP clients publish to social media directly. Add the server and start posting — zero configuration.
Skills
Install the Postproxy skill to give your AI assistant publishing capabilities. One command to add, then publish from your project without leaving the editor.
Agentic workflows
Connect Postproxy to n8n, Zapier, Make, or any workflow tool. Schedule posts, trigger on events, or let your automations decide when and what to publish.
One connection. Your agent publishes everywhere.
Wire Claude, ChatGPT, Cursor, Codex, Hermes, or OpenClaw into Postproxy once: publish out, pull insights back, handle comments and DMs as events — while your credentials stay server-side.
