Most people already draft their ad copy in Claude, argue about positioning in Claude and paste competitor screenshots into Claude. What they cannot do is act: Claude can describe a good TikTok hook, but it cannot pull the ad library, it cannot render the video, and it cannot post it. That gap is why marketing work still bounces between six tabs.
Hermoso closes it with a Model Context Protocol (MCP) server. MCP is the open standard for giving an assistant real tools, and Hermoso’s exposes 301 of them — ad-library research, creative generation, publishing, scheduling, campaign management and reporting — over a single connection you approve once. After that, marketing work happens where the thinking already happens.
Install it once: in Claude Code, Cursor or Codex run claude mcp add --transport http hermoso https://app.hermoso.ai/mcp. In Claude.ai or Claude Desktop, paste https://app.hermoso.ai/mcp under Settings → Connectors and approve it with your Hermoso account. For a terminal agent, npm install -g hermoso then hermoso auth login. Full setup on the MCP & CLI page.
What Claude can actually do once it’s connected
These are the six families of tool, and the plain-language ask that reaches each one.
| You say | What runs |
|---|---|
| “Who competes with oatly.com, and what are their longest-running Meta ads?” | find_competitors, pull_competitor_ads, competitor_teardown, mine_angles — plus the Meta, Google and LinkedIn ad libraries and organic TikTok, Instagram, YouTube, Reddit and Threads. |
| “Learn my brand from my website.” | draft_brand reads your site into a brand profile — products, palette, logo, voice — which every later render is grounded in. |
| “Make a 9:16 UGC video ad from that hook, my real product, captions burned in.” | plan_ad then render_ad / generate_video / generate_avatar, routed across 50+ image and video models. You get an mp4 URL back, not a description of one. |
| “Post it to Instagram, and schedule the cutdown for Tuesday.” | post_to_meta, post_to_tiktok, post_to_youtube, post_to_x, post_to_linkedin_page, post_to_pinterest, schedule_post — publish to eight channels, with one scheduling queue across Facebook, Instagram, Threads, TikTok and YouTube. |
| “Build a paused Meta campaign around the winner at $30/day.” | create_meta_campaign, create_meta_adset, create_meta_ad — and the Google Ads equivalents. Created paused; nothing spends until you confirm. |
| “Put last month’s results in a sheet.” | meta_insights, google_ads_report, then create_sheet / append_to_sheet in your own Google Drive. |
Why an MCP beats copy-pasting into a chat
Three things change when the assistant holds the tools rather than the transcript.
- It works from live data, not your description of it. Ad-library results are queried at the moment you ask, with real run dates, so “longest-running” means something. Nothing is reconstructed from memory.
- It returns artifacts, not prose. Every tool hands back a durable asset URL or a structured result. A render is an mp4 you can open; a campaign is an object id you can look up in Ads Manager.
- Context carries across steps. The research that produced the hook is still in play when the ad gets written, and your brand profile is still in play when it gets rendered. That continuity is exactly what breaks when you shuttle screenshots between four SaaS tools.
Brand memory is the part that compounds
A one-off prompt gets you a one-off ad. Hermoso keeps a brand profile — your products, real product photography, palette, logo and voice — plus a memory of what you told it and a library of what it made. remember, list_memory and use_brand mean a fresh chat on a fresh machine still knows your brand, and agencies can hold a separate brand-locked workspace per client with list_brands and use_brand. Skills you save with save_skill become reusable creative directives the agent applies next time.
What it will not do without asking
An agent with a budget is only useful if the blast radius is bounded, so spend authority is fenced in the server rather than in a prompt. Meta campaigns, ad sets and ads are created PAUSED with no caller override. The seven tools that can arm real ad spend — set_meta_campaign_status, set_google_ads_status, set_microsoft_ads_status, set_reddit_ads_status, set_linkedin_ads_status, set_pinterest_ads_status and set_openai_ads_status — each refuse to run without an explicit confirmation flag, and that same gate covers creating an enabled Google Ads ad group, ad or keyword under an already-enabled parent, because that serves on the very next auction. Deletes are confirm-gated too. On the generation side, hermoso_capabilities publishes each model’s exact credit cost before anything runs, and every render reserves an estimate then settles the true cost, so a failed dispatch refunds instead of quietly billing you.
You don’t have to give up the interface
Everything the tools do also exists as a visual studio at app.hermoso.ai — same engine, same credits, same library. Plenty of people research and publish from Claude but want to eyeball a render frame by frame before it goes out, and that is a perfectly good way to work. The point of the MCP is that the studio stops being the only way in.
If you live in a terminal rather than a chat window, the same surface is available through Claude Code and the Hermoso CLI.
Frequently asked
Can Claude run my marketing?
Yes, once you connect the Hermoso MCP server. Claude gets 301 marketing tools: it can research competitor ads across the Meta, Google and LinkedIn ad libraries plus organic TikTok, Instagram, YouTube, Reddit and Threads; generate finished image and video ads grounded in your real product; publish to Facebook, Instagram, Threads, TikTok, YouTube, X, LinkedIn and Pinterest; and build and read campaigns on Meta, Google Ads, Microsoft Advertising, Reddit Ads, LinkedIn, Pinterest and ChatGPT Ads. Without a tool connection, Claude can only advise — it cannot pull live ad data, render a video, or post anything.
How do I connect Hermoso to Claude?
In Claude.ai or Claude Desktop, paste https://app.hermoso.ai/mcp under Settings then Connectors and approve it with your Hermoso account. In Claude Code, run: claude mcp add --transport http hermoso https://app.hermoso.ai/mcp. There is nothing to self-host and no API keys to manage — the connection is OAuth’d to your account.
Does Claude need my Facebook or Google Ads password?
No. You connect Facebook, Instagram, Threads, TikTok, YouTube, Google Ads and Google Drive to Hermoso through each platform’s own OAuth flow, in Hermoso’s Connectors settings. Hermoso holds the resulting tokens; Claude only calls Hermoso’s tools. No password ever passes through the model.
Can Claude spend my ad budget by accident?
No. Meta campaigns, ad sets and ads are created paused with no caller override, and the seven tools that can arm real ad spend — set_meta_campaign_status, set_google_ads_status, set_microsoft_ads_status, set_reddit_ads_status, set_linkedin_ads_status, set_pinterest_ads_status and set_openai_ads_status — each refuse to run without an explicit confirmation. Generation spends Hermoso credits from your balance, and each model’s exact per-render cost is published through hermoso_capabilities so the agent can quote you a number first.
What does it cost to use Hermoso in Claude?
The MCP server itself is free to install and signup is free with 250+ earnable free credits and no card. After that it is one credit system across research, image and video, with plans at $19, $49 and $149 a month; every feature is on every plan and plans differ only by monthly credits and per-credit rate.
Does it work in ChatGPT, Cursor or other agents too?
Yes — it is a standard MCP server, so it works in any MCP client. There are verified paths for Claude, Claude Code, Claude Desktop, Codex, Cursor, VS Code, Cline, OpenClaw and Hermes.
Create a free Hermoso account, paste one URL into Claude, and your next marketing task is a sentence. No card, 250+ earnable free credits.
Start free → Read the MCP docs →