@houtini/contentmarketingideas

The MCP server for Content Marketing Ideas — the editorial copilot I built because I was tired of the same content research loop. You read a lot, you notice patterns, you get ideas, you write. The reading and pattern-spotting is the slow part, and most of it is mechanical.

This server pulls all of it into Claude Desktop, Claude Code, Cursor, or any MCP client: source monitoring, editorial briefs, voice-matched article drafts, RAG corpus search, Google Search Console data, and WordPress publishing. Around 70 tools, one npx command.

I wrote a long post about why the product exists and how I use it every day: Content Marketing Ideas — what it is, how I built it, and why I use it.

Quick Navigation

Get started | What it does | Tools reference | Configuration | Requirements

Get started in two minutes

Step 1: Get an account and an API key

Sign up at contentmarketingideas.co. Free tier covers five sources and weekly delivery, which is enough to wire the MCP up and try every tool. Once you're in, go to Dashboard → Account → API & MCP Server and click Generate API Key. The full key is shown once — copy it now. Keys are prefixed cr_.

Step 2: Add to your Claude Desktop config

Config file locations:

• Windows: C:\Users\{username}\AppData\Roaming\Claude\claude_desktop_config.json

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

{
  "mcpServers": {
    "contentmarketingideas": {
      "command": "npx",
      "args": ["-y", "@houtini/contentmarketingideas"],
      "env": {
        "CMI_API_KEY": "cr_your_api_key_here"
      }
    }
  }
}

Step 3: Restart Claude Desktop

Tools show up automatically. npx pulls the package on first run — no separate install needed.

Claude Code (CLI)

Claude Code uses claude mcp add rather than the desktop config file:

claude mcp add -e CMI_API_KEY=cr_your_api_key_here -s user contentmarketingideas -- npx -y @houtini/contentmarketingideas

Verify with claude mcp get contentmarketingideas — you should see Status: Connected.

Cursor / Windsurf / VS Code

Same shape, different config file. Add this block:

{
  "contentmarketingideas": {
    "command": "npx",
    "args": ["-y", "@houtini/contentmarketingideas"],
    "env": {
      "CMI_API_KEY": "cr_your_api_key_here"
    }
  }
}

Local build instead

For development, or if you'd rather not rely on npx:

git clone https://github.com/houtini-ai/contentmarketingideas-mcp
cd contentmarketingideas-mcp
npm install
npm run build

Then point your config at the local build:

{
  "mcpServers": {
    "contentmarketingideas": {
      "command": "node",
      "args": ["C:/path/to/contentmarketingideas-mcp/dist/index.js"],
      "env": {
        "CMI_API_KEY": "cr_your_api_key_here"
      }
    }
  }
}

Related MCP server: Fourth Brain Demo

What it does

Content Marketing Ideas isn't a "here's what's trending" feed. It's an editorial copilot — every brief is framed as "here's what you should write and why" rather than a data dump. The MCP exposes the same engine to Claude.

Editorial briefs from your sources

Use contentmarketingideas:list_briefs to show me this week's briefs for site-ae01a256

The product watches RSS feeds, YouTube channels, subreddits, podcasts, and news keywords for you. When something relevant lands, four signal generators (trending, decay, untapped, gap) write briefs containing a thesis, editorial play, gap analysis, suggested format, what to cover, internal link suggestions, and two alternative angles (obvious / bold / contrarian). Each brief is enriched with keyword volume, difficulty, monthly trend, People Also Ask questions, competitive landscape, entity relationships from Wikidata, and an originality score versus existing coverage.

get_brief returns the full payload — synthesis, signal sources, enrichment, momentum sparkline, the lot. brief_feedback lets you train the system from inside Claude (thumbs_up / thumbs_down / made_this / already_covered). Thumbs-down auto-archives briefs you don't want to see again.

Voice-matched article generation

Use contentmarketingideas:create_article with source_type="brief" and source_id="<brief_id>"

The web app analyses your published content with sixteen statistical analysers and produces a deterministic ~17KB writing-style guide. Article generation feeds that guide into Claude Sonnet so the output reads like you wrote it — not like an LLM. 1,300–1,800 words, full markdown, with Schema.org provenance and entity-tagged facts traceable back to the source.

generate_article queues an article for the production pipeline, regenerate_article re-runs failed jobs, get_article_provenance returns the source chain and Schema.org JSON-LD, and publish_article pushes the finished post straight to WordPress over the REST API (credentials encrypted at rest).

Topic Explorer — coverage as a map, not a list

Use contentmarketingideas:get_coverage_explore for site-ae01a256

Topics are the unit of value. The Topic Explorer assembles every entity from your last 100 briefs plus your site's core topics, then enriches each with corpus matches, GSC traffic and 28-day sparklines, search volume, competitive landscape, parent/sibling topics from Wikidata, and a lifecycle state: covered, trending, gap, declining, or emerging. get_coverage_explore returns the full clustered dataset so you can ask Claude things like "which trending topics do I have zero coverage on?" or "cluster my declining pages by parent topic and suggest refreshes."

Search Console integration

Use contentmarketingideas:gsc_growing_queries with site_id="..." and limit=20

GSC OAuth is per-site. Once connected, daily data feeds into brief generation and is queryable through the MCP: gsc_summary, gsc_queries, gsc_opportunities (high-impression, low-position), gsc_growing_queries, gsc_declining_queries, gsc_growing_pages, gsc_declining_pages, gsc_daily_summary for sparklines.

RAG corpus — your own knowledge base inside Claude

Use contentmarketingideas:search_rag with q="long-tail audience research methods" and limit=10

The corpus is human-curated by design — your own published content, the sources you follow, and URLs you add manually. No auto-crawling junk. Documents are indexed into Cloudflare Vectorize (bge-m3, 1024d) and a D1 FTS5 index for hybrid semantic + keyword search. Long structured documents get section-aware embeddings with parent_id linkage, and queries dedupe sections back to their parent doc with a small breadth bonus.

populate_corpus indexes a list of URLs, populate_site walks a sitemap, populate_youtube pulls full transcripts from a YouTube channel. list_corpus_documents, delete_corpus_document, rag_stats, and search_rag round it out. Site-scoped — sources for site A don't pollinate site B's briefs.

Source management

Use contentmarketingideas:detect_source with url="https://example.com/feed"

detect_source auto-identifies feed type (RSS, YouTube, Reddit, podcast, news) and resolves the canonical feed URL. add_source, update_source, delete_source, list_sources, get_source cover the rest. Sources are site-scoped, so a multi-site account never cross-contaminates briefs.

Pipeline triggers

Use contentmarketingideas:trigger_synthesis to generate briefs now

Don't want to wait for the cron run? trigger_ingestion runs all your sources, trigger_synthesis generates briefs from current content, trigger_profile re-runs site profile analysis (voice guide, topics, audience, content gaps). Useful for testing and for end-of-day "what came in today" runs.

Account, sites, deliveries, WordPress

Multi-site accounts are first-class. list_sites / create_site / update_site / delete_site / update_site_schedule for site management; list_account_emails / add_account_email / delete_account_email for multi-recipient delivery (Business+); get_wordpress_status / save_wordpress_credentials / delete_wordpress_credentials for publishing; list_deliveries for the email delivery audit trail.

Tools reference

Around 70 tools across seven groups.

Briefs

Sources

Articles

RAG corpus

Growth & GSC

WordPress

Account, sites, pipeline

Configuration reference

Security

Your API key authenticates as your account and your account only. Every request is server-side authenticated against users.api_key (hashed). The API enforces ownership checks on every resource: you cannot read, modify, or delete another user's data even by guessing IDs. No X-User-ID impersonation header exists.

If a key leaks, revoke it from Dashboard → Account → API & MCP Server and generate a new one — old keys are immediately rejected.

Plans and limits

The MCP itself is free and open source. The platform behind it has tiered limits:

Free is enough to wire up the MCP and try every tool against real data. Pro unlocks daily delivery, voice-matched drafts, growth plans, and keyword analytics. Business adds multi-site, WordPress publishing, and competitor intel. Pricing is at contentmarketingideas.co/pricing.

Requirements

• Node.js 18+

• A Content Marketing Ideas account (free tier works) and an API key

• An MCP-compatible client: Claude Desktop, Claude Code, Cursor, Windsurf, or any other MCP host

Links

• Product: contentmarketingideas.co

• Why I built it: houtini.com — Content Marketing Ideas: what it is, how I built it, and why I use it every day

• npm: @houtini/contentmarketingideas

• MCP spec: modelcontextprotocol.io

Licence

MIT