Viral Cat in Claude

Connect Viral Cat to Claude Desktop and run your social posting by asking for it. No dashboard, no clicking around. Ask what's coming up, ask why something failed, ask for a draft, and Claude works straight from your real Viral Cat account.

Connect it

In Claude Desktop

Open Settings, go to Connectors, and choose Add custom connector. Paste this address and your Viral Cat token as a Bearer token.

https://viral-cat.com/api/mcp/mcp

Using Claude Code or another stdio client instead? Add this to your MCP config.

{
  "mcpServers": {
    "viral-cat": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://viral-cat.com/api/mcp/mcp",
        "--header",
        "Authorization: Bearer YOUR_TOKEN"
      ]
    }
  }
}

Your token is the key to your account. Keep it private, never paste it into a shared chat or a public repo, and generate a fresh one at viral-cat.com if you think it has leaked.

Just ask

Plain English is the whole interface. Some things people ask:

What's queued to post this week?

Which posts failed, and why?

Draft a post about our summer sale for my Acme brand.

Reschedule Friday's post to Monday morning.

Is my Threads connection still healthy?

What it can do

The boundaries are deliberate. Here's both sides of them.

It can

  • +Read your brands, posts, and everything sitting in the pipeline
  • +Check whether each platform connection is still healthy
  • +Pull performance numbers for posts that already went out
  • +Explain why something failed, in plain English
  • +Write drafts for you to look over
  • +Schedule posts for a future date and time
  • +Cancel or move a post that hasn't gone out yet
  • +Retry failures that look recoverable
  • +Refresh a platform token that has gone stale

It can't

  • ×Publish something right now. Everything goes into the queue with a minimum lead time before it can go out
  • ×Show you, or hand over, your platform tokens
  • ×Post to an account you didn't name

Every write asks first

Nothing changes in your account without you clicking to allow it.

All six of the actions that change something stop and ask you in Claude before they run. You see exactly what is about to happen and confirm it yourself. Say no and nothing is written at all, not a draft, not a schedule change, nothing.

Retrying failures is extra careful: it does a dry run by default, so you get a list of what it would retry before anything is actually re-queued.

Everything it can reach

The complete list, nothing hidden. Reads happen freely; writes ask first.

Things it can look at

Your brands

list_brands

Every content profile you run, with its post counts.

Brand details

get_brand_details

The full settings and posting history for one brand.

Your posts

list_posts

Browse posts across all platforms, newest first.

Pipeline summary

get_pipeline_summary

A quick snapshot of what's drafted, queued, and done.

Platform connections

list_connections

Which accounts are linked and whether they still work.

Post performance

post_performance

How your published posts actually did.

Content queue

content_queue

What's coming up next, in order.

Campaign summary

campaign_summary

A per-brand rollup of posted, scheduled, and failed.

Failure diagnosis

diagnose_failures

Why something didn't post, and what would fix it.

Preflight check

preflight_check

Catches problems before a post is due to go out.

Things it can change, with your say-so

Write a draft

create_draft

Saves a draft for you to review. Nothing is published.

Schedule a post

schedule_post

Queues a post for a future date and time you choose.

Cancel a post

cancel_post

Pulls a queued post back out before it goes anywhere.

Retry failures

retry_failed_posts

Re-queues failures that look recoverable.

Move a post

reschedule_post

Shifts a queued post to a different day or time.

Reconnect a platform

reconnect_platform

Refreshes a platform token that has gone stale.

Prefer to wire it up yourself? The same actions are available over plain HTTP in the API reference.