Run your whole social presence from Claude: 281 tools, two setup paths

As of this summer, the Claude integration covers the entire product. Not "the scheduling parts." All of it. The number under the hood is 281 tools, which is a number I can only make meaningful by telling you what it feels like to use.

I've written before about why conversation beats automation and what a Cowork session looks like minute-by-minute. This post is the other half: the full capability map, the two ways to connect, and the safety rules that make it something you can trust with your accounts and your money.

What 281 tools actually means

Nobody needs a list of 281 anything, so here's the practical translation. Each of these is one conversation:

"Schedule my week." Claude reads your books library and brand voice, drafts platform-specific captions, generates images in Media Studio, and lays the posts into your queues. You review, tweak two captions, approve. Twenty minutes of clicking becomes four minutes of reading.

"Draft an ad campaign for Book 3, I'll review before anything runs." The ads tools are all there: Claude proposes audiences and hooks grounded in your brand guide, checks the targeting against the network's real interest search, and hands you a draft. It can also run the Optimize analysis and tell you which campaigns earned a scale, hold, trim, or kill verdict this month.

"Triage my inbox." Claude pulls your unread DMs, comments, and reviews, summarizes what needs a human, drafts replies for your approval, and can build or adjust the automation flows that handle the routine questions.

"How did July go?" The full analytics surface is queryable: follower growth, best posting times, which posts decayed fastest, whether anything failed to publish and why.

"Make me a launch graphic." Media Studio generation works from chat, grounded in the same brand guides your dashboard uses, so the output looks like yours.

The shape underneath: if you can do it in the dashboard, you can now ask for it in a sentence. Same account, same data, same rules.

Two ways to connect

Everything lives on one card in Settings: "Use Author Automations inside Claude." It walks you through creating a credential, then splits into two tabs, because there are two kinds of Claude user and they need different last steps.

Path A: the Claude.ai connector. Pick this if you use Claude in the browser or the desktop app, including Cowork. There's nothing to install. The card gives you a remote server URL to paste into Claude's connector settings along with your credential, and from then on every conversation can see your account. The server runs on our side; your laptop is not involved.

Path B: the Claude Code plugin. Pick this if you live in the terminal. Install the plugin, sign in once, and the same 281 tools show up in your Claude Code sessions, right next to whatever else you work on there. Same capabilities as the connector, different home.

Both paths use the same credential pair from the card, and you can revoke any credential from the same place at any time. The whole setup is five minutes, and the card holds your hand through every step of both tabs.

Pen names are respected everywhere

If you write under more than one name, this matters more than any feature: when you create a credential, you choose which pen names it can reach. Give Claude access to all of them, or scope a credential to just one.

And in conversation, every tool operates within a pen name's walls. Ask about your thriller pen name's calendar and Claude sees that calendar, those accounts, those books, that brand voice. The separation you rely on in the dashboard doesn't loosen because you came in through a conversation. Cross-pen-name work is explicit: you say which name you're working under, or you ask Claude to switch.

The safety posture, spelled out

This is the section I'd want to read before connecting an AI to accounts that can spend money, so here it is plainly.

Money actions confirm, every time. Any tool that touches real spend, launching a campaign, resuming a paused one, boosting a post, sending a broadcast to real contacts, is built to state the budget and duration and get your explicit yes first. That rule is written into the tools themselves, not left to the model's mood.

Launches default to paused. When Claude creates a boost or a campaign, it's created paused unless you explicitly say "launch live." Amazon campaign creation goes one step further: it is always created paused, and you activate it in Amazon's own console after review. Drafting is free and commits nothing.

Analysis never applies itself. The ads Optimize engine returns verdicts and suggested actions. Acting on them is a separate step that you approve.

Destructive actions are guarded. Permanent deletes and bulk edits require you to explicitly confirm what's changing before anything is touched.

The design principle underneath: Claude drafts, proposes, and reads freely, and asks before anything irreversible or expensive. You should be able to hand this to your most cautious self and have it hold up.

Where this fits in a real week

The dashboard doesn't go away. Visual work, calendar drag-and-drop, one-off typo fixes, all still faster with a mouse. What changes is the sessions that used to be sessions: the Sunday scheduling block, the launch-week scramble, the monthly analytics review, the "I should really deal with my inbox" dread. Those compress into conversations you can have from any Claude surface, in the gaps of a writing day.

My honest pitch is the same one I made when this was a much smaller plugin: the tool finally meets you where your attention already is. Except now it's not just posting. It's the whole practice.

What to do next

• Open Settings, find the "Use Author Automations inside Claude" card, and pick your tab. Connector if you chat, plugin if you terminal.
• Start with a read-only question: "What's on my calendar this week?" Prove to yourself it sees your account before you ask it to change anything.
• Then try one real task end to end. My suggestion is inbox triage, because it's the one where four minutes of conversation most obviously replaces an hour of app-hopping.