MCP Server

What it does

One command turns all 17 CLI tools into MCP (Model Context Protocol) tools. Your AI agent — DeepSeek Harness, Claude Desktop, Codex, or any MCP client — can call them directly. You describe the task in plain language; the agent picks the tool, fills the arguments, and reads the output.

python3 -m zens_ink.mcp

Or via the CLI entry:

zens-ink mcp

Tool names

Each CLI tool is exposed as mcp__zensink__<tool_name>:

• keyword_research, keyword_volume, brave_volume
• kd, serp_intent, search_intent
• keyword_cluster, kgr_auto, content_matrix
• search_performance, competitor_gap
• site_audit, onpage_audit, geo_fanout, reddit_blueocean, rank_tracker

Client configuration

DeepSeek Harness (DSH) — add to your DSH profile:

{
  "mcpServers": {
    "zensink": {
      "command": "python3",
      "args": ["-m", "zens_ink.mcp"],
      "cwd": "/path/to/zens-ink-seo-package"
    }
  }
}

Claude Desktop — claude_desktop_config.json:

{
  "mcpServers": {
    "zensink": {
      "command": "python3",
      "args": ["-m", "zens_ink.mcp"],
      "cwd": "/path/to/zens-ink-seo-package"
    }
  }
}

API keys are read from the package root .env — same file as CLI usage. No keys, no problem: key-free tools (keyword_research, keyword_cluster, site_audit, onpage_audit) work without any configuration.

Pro auto-discovery

If the Pro package (zens_ink_pro) is installed in the same environment, its tools — winability, content_radar, competitor_radar, geo_score, geo_visibility, gap_deep, full_audit, content_briefs, ai_sov, agent_readiness — join the same MCP server automatically. OSS-only installs stay OSS-only.

Example prompts

Once connected, just ask:

Find me long-tail keywords for "astro templates", then cluster them

What's the keyword difficulty for "seo cli tool"? Check if it's a brand keyword first

Audit my dist/ folder for technical SEO issues