How to Schedule Social Media Posts with Claude (Yes, Actually Publish Them)

Ask Claude to write you a week of Instagram captions and it will, happily and well. Ask it to actually put those captions on a schedule, and by default it can't. Claude on its own has no hands. It can't reach out to Instagram's servers, it can't hold your access tokens, and it has no idea what your account's engagement history looks like.

Connect it to an MCP server built for social media and that changes. The drafting stays the same, but now there's a second half: scheduling, publishing, and analytics, done by the same agent, in the same conversation. One scope note before the walkthrough: drafting works across every platform TimeToPost supports, but live scheduling and publishing through the connector currently reaches X; Instagram, TikTok, Facebook, and Threads are coming soon. This is the practical guide to setting that up and using it, not the theory.

What Claude can do natively vs with a connector

Native Claude, no connector, is a writer. Give it your brand voice, your last ten posts, a topic, and it will produce captions, hooks, hashtags, and alt text that sound like you. It's genuinely good at this. What it cannot do without help is anything that requires touching the outside world: no posting, no scheduling, no reading your account's real metrics.

A remote MCP connector closes that gap. MCP (Model Context Protocol) is the open standard that lets Claude call out to a tool server over HTTP, the same way it might call a search tool or a code sandbox. A social scheduling MCP server exposes tools like "create a draft," "schedule a post," "publish now," "get last week's engagement." Claude decides when to call them based on what you ask for. The model doesn't get smarter about your brand, it just gets an actual set of hands.

Anthropic's connectors directory has grown past 400 verified integrations as of this writing, spanning everything from project management to CRMs to, yes, social scheduling. TimeToPost runs a hosted one at https://api.timetopost.co/mcp with tools for drafts, scheduling, publishing, threads, analytics, optimal posting times, and approvals.

Two-minute setup (Claude Desktop, Claude Code, claude.ai)

Remote MCP connectors work the same way across Claude's surfaces, which is the whole point of the standard. You're not configuring three different integrations, you're configuring one and using it from wherever you happen to be working.

In Claude Code, the fastest path is the CLI:

claude mcp add --transport http timetopost https://api.timetopost.co/mcp

That registers the connector for your Claude Code sessions. The first time Claude tries to use a TimeToPost tool, you'll get a browser window asking for your email, and then a one-time code. No password ever touches Claude, no API key sits in a config file you have to remember to rotate. Approve it once and it's live.

Claude Desktop and claude.ai both support the same remote connector under Settings, usually under a "Connectors" or "Integrations" section, where you paste the same URL. The auth flow is identical: browser, email, code.

If you ever want it gone, it's gone in one click. Revoke access from Settings -> API on the TimeToPost side and the connector loses its token immediately, no email to support required.

Workflow 1: draft a week of posts in one conversation

This is the workflow most people start with, because it's the one closest to what Claude already does well. Open a conversation and say something like: "Draft five posts for this week about our new pricing page, one for X, one for LinkedIn, three variations for Instagram." Claude writes the copy the way it always has. The difference now is what happens next.

Instead of copying each draft into five browser tabs, ask Claude to save them. With drafting tools available, it can create each post as a draft directly in your TimeToPost queue, tagged and dated, ready for you to review. You stay in one conversation the whole time: write, revise, save, repeat, until a week's worth of content exists somewhere other than a scratch document you'll lose track of by Thursday.

Workflow 2: schedule at your best times automatically

Drafting is the easy half. Scheduling well requires knowing something Claude doesn't have by default: when your specific audience is actually online and engaging. That's where the optimal-times tool matters. Claude can call get_optimal_times against your connected account's own history, not a generic industry benchmark, and place each draft on a real slot based on when your last few months of posts actually performed.

In practice this looks like asking, "schedule these five for whenever they'll do best this week," and Claude fills in the calendar using your account's own numbers instead of guessing at 9am and 5pm because that's what every blog post says. If you'd rather see the logic without touching a chat window, TimeToPost's best-time-to-post calculator runs the same underlying analysis on demand.

Workflow 3: ask for last week's numbers

The third workflow closes the loop, and it's the one people forget is possible until they try it. Once posts have actually gone out, ask Claude "how did last week's posts do, and which platform is pulling its weight." With analytics tools connected, Claude can pull real engagement numbers per post and per platform and summarize them in plain language instead of a spreadsheet you have to open and squint at.

This matters more than it sounds like. Most people who "check their analytics" mean opening a dashboard once a month and skimming it. Being able to ask a question in a sentence and get a straight answer is the difference between analytics you glance at and analytics that actually change what you post next week.

Skills and standing instructions (voice rules, approval rules)

The workflows above assume Claude knows your rules every single time, which it will not do from memory alone. This is what Claude Skills are for: a reusable set of instructions you write once (your brand voice, banned words, hashtag limits, which platforms get which tone) that Claude applies automatically to relevant requests without you repeating yourself in every conversation. Skill availability can vary by plan, so check what your account has access to before you build a workflow that depends on it.

The same principle applies to approvals, and it matters more here because the stakes are higher than a poorly worded caption. TimeToPost's MCP server exposes approval-queue tools, so you can set up a standing instruction like "draft everything, but nothing publishes until I approve it in the queue." Claude does the writing and the scheduling logic, a human still says yes before anything goes live. That's a rule you set once and it holds for every conversation after.

Security model: what the agent can and cannot touch

The honest question anyone should ask before connecting an AI agent to their social accounts is: what happens if it gets something wrong? The answer depends entirely on the access model, and a well-built connector is deliberately narrow.

Claude never sees your Instagram password, your X password, or any platform credential directly. The browser-based email-code authentication issues a scoped, revocable token to the connector, and that token is what Claude's tool calls use. You can revoke it in Settings -> API at any moment, and the connection dies immediately, no residual access, no lingering session.

Within that token, the tools available define the ceiling. If you only want Claude drafting and reading analytics, you can leave publishing tools out of the loop entirely. If you want it to schedule but never publish without a human clicking approve, the approval queue tools enforce that. The agent can only do what the tool surface lets it do, and you control that surface. This is not "give an AI your login," it's closer to giving a very fast intern a specific set of keys and a lock on the register.

FAQ

Do I need a paid Claude plan to use MCP connectors? MCP connector support itself is broadly available across Claude Desktop, claude.ai, and Claude Code, but individual features like Skills can be gated by plan tier, so check your account's current entitlements before building a workflow around them.

Can Claude post without my approval? Only if you let it. With approval-queue tools in place, drafts sit until a human approves them. Whether Claude publishes directly or waits for sign-off is a setting you choose, not a default you're stuck with.

Does Claude need my social media passwords? No. Setup uses browser-based email and one-time-code authentication, and the resulting access is scoped and revocable from Settings -> API at any time. Your passwords never enter the conversation.

What platforms can Claude actually publish to through this? Through a connector like TimeToPost's, native publishing and scheduling cover X today, with analytics reporting for whatever you've connected. Instagram, TikTok, Facebook, and Threads publishing are coming soon; drafting works across all of them right now.

Is this different from just asking ChatGPT to write my posts? Writing captions is something most capable models can do. The difference here is the connector layer: tools that let Claude actually schedule, publish, and read your account's real analytics instead of stopping at a block of text you still have to move by hand.

What if I only want drafting, not publishing? That's a reasonable and common setup. Give the agent drafting and analytics tools and leave publishing out entirely, and Claude will hand you finished drafts in the queue for you to schedule yourself.

If you want the deeper explanation of what MCP servers for social media actually are and aren't, see Social Media MCP Servers, Explained. And if the bottleneck isn't the posting but the drafting itself, batching a month of content in one afternoon pairs well with handing the resulting queue to Claude.