PostLake — MCP server & agent skills for social media

PostLake is the social media API for AI agents. One integration to publish, schedule, and measure across every network — X, LinkedIn, Instagram, TikTok, Facebook, Threads, Bluesky, YouTube and Pinterest — with one normalised response shape an agent can reason over instead of nine different error formats.

This repo holds the agent skills and documents the hosted MCP server.

• 🔌 MCP server: https://api.postlake.dev/mcp (streamable-http, OAuth)
• 📚 Docs: https://postlake.dev/docs/
• 🌐 Site: https://postlake.dev
• 📖 For LLMs: https://postlake.dev/llms.txt

MCP configuration

PostLake is a remote MCP server — there is nothing to install, run, or self-host. Point your client at the URL and approve once over OAuth:

{
  "mcpServers": {
    "postlake": {
      "url": "https://api.postlake.dev/mcp"
    }
  }
}

Note: PostLake is not launched with a local command. Any config that runs npx to start this server is incorrect — the server is hosted at the URL above. (The npx skills add … command further down installs the optional agent skills; it does not start an MCP server.)

Quick start — MCP

Your agent never sees an API key; OAuth handles authorisation.

Claude

claude mcp add postlake https://api.postlake.dev/mcp

Or in Claude Desktop: Settings → Connectors → Add custom connector.

Cursor

~/.cursor/mcp.json:

{
  "mcpServers": {
    "postlake": { "url": "https://api.postlake.dev/mcp" }
  }
}

Gemini CLI

gemini mcp add postlake https://api.postlake.dev/mcp

VS Code / GitHub Copilot

Agent mode → MCP servers → Add, then paste the same URL.

Header auth (for clients without OAuth)

POST https://api.postlake.dev/v1/mcp
Authorization: Bearer <your_api_key>

Then just ask:

"Post this photo to Instagram and LinkedIn, schedule a follow-up for Friday at 9am, and tell me which of last week's posts did best."

The toolset

A focused set — everything an agent needs to publish and reason about the result, and nothing it doesn't. Account setup and connecting networks stay in the dashboard, where a human belongs.

Agent-safe by design. Publishing is idempotent — pass an Idempotency-Key and a retrying or restarting agent can never double-post. Every post returns a per-platform state and url in one shape, so there's no per-network branching.

Agent skills

For coding agents (Claude Code, Cursor, Codex, Windsurf) that use file-based skills rather than MCP:

npx skills add postlake/postlake-mcp --all

Set your key once:

export POSTLAKE_API_KEY="sk_live_…"

Create one at https://app.postlake.dev/app/keys.

Three ways to connect

REST base URL: https://api.postlake.dev/v1 — see the full docs and per-framework guides for LangChain, CrewAI, AutoGen, Python and Node.js.

Security

Agents authorise over OAuth, so an API key is never pasted into a chat or seen by the model. Every agent you authorise appears under Connected Apps in your dashboard — revoke any one and its access is cut off immediately.

Pricing

Usage-based credits: 1 credit per post, with a free tier and no card to start. X costs more because it's the only network that charges per post. See https://postlake.dev/pricing.

Listed in the official MCP Registry as dev.postlake/social.

Made by CrumbleLake. Issues and PRs welcome.