What is MCP?
The Model Context Protocol is an open standard letting AI assistants (Claude Code, Cursor, others) access external tools and data sources. Rankion provides an MCP server exposing every API endpoint as a "tool" for the assistant.
Create API token in Rankion
In Rankion, go to Settings → API & Integrations → "+ New Token". Give a meaningful name (e.g. "Claude Code Mac"), select scopes (full access: "all"), and copy the token immediately — shown only once.
Heads up
Treat the token like a password. Whoever has it can access your data AND spend credits. Never push to public repos.
Configure MCP server in Claude Code
Open your Claude Code MCP config (`~/.claude/mcp_servers.json` or via `claude mcp add`). Add this entry:
{
"mcpServers": {
"rankion": {
"command": "npx",
"args": ["-y", "@rankion/mcp-server"],
"env": {
"RANKION_API_TOKEN": "rk_live_xxxxxxxxxxxxxxxxxxxxxx"
}
}
}
}Alternatively as one-liner via the Claude CLI.
claude mcp add rankion \
--command "npx -y @rankion/mcp-server" \
--env RANKION_API_TOKEN=rk_live_xxxxxxxxxxxxxxxxxxxxxxTest the connection
Restart Claude Code. In any session, type.
Liste meine Rankion-Projekte aufIf the connection works, Claude responds with your project list. If not: check token (valid?), JSON syntax (commas?), and that `npx` is in your `$PATH`.
First real commands
Three examples that show what MCP is really useful for.
# Keyword-Recherche aus dem Terminal
> Mach eine A-Z-Expansion für "nachhaltige rucksäcke" im Projekt "Outdoor"
und gib mir die Top-20-Keywords nach Suchvolumen.
# Artikel direkt erstellen lassen
> Generiere im Projekt "Outdoor" einen Artikel für das Keyword
"wanderrucksack 30l damen test 2026" und exportiere ihn als Markdown
in ./content/.
# AI-Visibility-Report on demand
> Wie ist meine Mention Rate bei "beste seo tools 2026" diese Woche
in ChatGPT und Perplexity? Vergleiche mit letzter Woche.More complex workflows
The real lever: Claude chains MCP tools autonomously. You set a goal, Claude orchestrates the steps. Example.
> Analysiere das Top-3-Keyword im Projekt "Magazin" auf AI-Sichtbarkeit,
finde drei Wettbewerber, die besser ranken, identifiziere die fehlenden
Subtopics und schlage drei neue Artikel vor mit Briefings, die ich
direkt in Bulk Generation kippen kann.Result
Result: An SEO co-pilot directly in your terminal that doesn't just fetch data but actively suggests and executes strategies — without ever opening the web UI.
What you learned
- ✓ Understood MCP as open standard for connecting AI assistants with tools
- ✓ Created API token with correct scope and stored securely
- ✓ Configured MCP server entry in `~/.claude/mcp_servers.json`
- ✓ Verified connection with "list my projects" command
- ✓ Orchestrated multi-step workflows (research → comparison → briefs) via natural language