ReplyNodesReplyNodes← All posts

Connecting Claude Code to Your Social Channels

June 8, 2026 · ReplyNodes Team

Claude Code spends most of its day in your codebase — but a lot of what teams ship deserves an announcement: a new feature, a changelog entry, a release. Instead of context-switching to a social scheduling tool, you can let Claude Code draft and schedule that announcement itself, using the same MCP connection it already uses for everything else.

What the integration looks like

ReplyNodes exposes an MCP server with the same tools every agent uses: drafting per-channel content, listing connected accounts, requesting approval, and scheduling. Once Claude Code is connected, a prompt like:

"We just shipped the new approval workflow feature. Draft an announcement for X and LinkedIn, and put it up for approval — don't auto-publish."

...is enough. Claude Code reads the relevant commit or changelog for context, drafts channel-appropriate copy (X gets something tighter than LinkedIn), and submits it through ReplyNodes' approval queue rather than publishing directly — because that's how the tool is scoped, not because the prompt asked nicely.

Why route it through approval by default

This is the part that matters most for teams actually shipping this in production. An agent that can publish unsupervised is a liability the first time it misreads context. ReplyNodes' approval workflows (see Approval workflows for AI-generated social content) let you decide per-channel whether agent-drafted posts need a human sign-off or can go straight out — Claude Code doesn't get to make that call itself, the workspace configuration does.

Setting it up

  1. Generate an MCP connection token from your ReplyNodes workspace settings.
  2. Add the ReplyNodes MCP server to Claude Code's configuration, same as any other MCP server.
  3. Confirm which channels are connected and which require approval — this is workspace-level, not something Claude Code needs to be told per request.
  4. Start drafting. Claude Code will use list_channels, draft_post, and request_approval as needed based on what you ask for.

Full setup steps and the tool reference live in the documentation and API reference. If you're comparing this to a raw API integration, see Agent-native vs. API-first: what's actually different for why the MCP layer changes what's actually safe to automate.