Skill for Claude Code (instead of an MCP server)

Important — clarification 2026-05-16: Rankion does not run an MCP server. Earlier versions of this wiki announced a planned MCP endpoint at /mcp — that plan has been dropped. Instead we ship a Skill (a single Markdown file for Claude Code), our REST API with 130+ endpoints, and an OpenAPI spec for every other agent.

rankion.aiWhat you get instead

1. Skill for Claude Code — a Markdown file at https://rankion.ai/claude-skill that Claude Code reads directly. No npm package, no MCP server, no extra infrastructure.
2. REST API — 130+ endpoints under https://rankion.ai/api/v1/..., documented in the API reference. Auth via Sanctum bearer tokens.
3. OpenAPI spec — machine-readable at https://rankion.ai/api/openapi.json. Use it to auto-generate type-safe clients for any agent framework (n8n, Zapier, ChatGPT connector, custom tooling).

A single Sanctum token works for all three paths.

rankion.aiSetup for Claude Code in 2 minutes

# 1. Create the skill directory
mkdir -p ~/.claude/skills/rankion-api

# 2. Pull the skill Markdown from rankion.ai
curl -o ~/.claude/skills/rankion-api/SKILL.md https://rankion.ai/claude-skill

# 3. Set your Sanctum token (copied from Settings → API)
export RANKION_API_TOKEN="rk_live_..."

# 4. Launch Claude Code — the skill is available immediately
claude "Research keywords for 'AI Content Marketing' and draft an article"

Create the token in the UI under Settings → API.

rankion.aiFor other agents (ChatGPT connector, Cursor, n8n, Zapier, custom)

Every non-Claude agent calls the REST API directly. Example:

curl -X POST https://rankion.ai/api/v1/generate/article \
  -H "Authorization: Bearer $RANKION_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"project_id": 1, "keyword": "AI Content Marketing", "article_type": "blog-post"}'

For frameworks that understand OpenAPI: load https://rankion.ai/api/openapi.json and auto-generate type-safe clients.

rankion.aiIf your agent really requires MCP

As of 2026-05-16 we have not encountered such a case. Current MCP clients (Claude Desktop, Cursor, Windsurf) can in nearly every setup call REST APIs directly — or, in the case of Claude Code, load a skill. If your concrete use case strictly needs an MCP server, email info@provimedia.de — we evaluate case by case; there is no roadmap commitment.

rankion.aiBest practices

• Never commit the Sanctum token. Keep it in .env or inject via a secret manager.
• One token, one team. Sanctum tokens are team-scoped — they see only the team that owns them.
• Watch rate limits. The Skill and OpenAPI doc list per-endpoint rate limits (typically 60/min).
• Poll async endpoints correctly. Long-running operations (article generation 30-90 s, audits) return HTTP 202 with a job ID — poll the GET status endpoint until status: "completed".

rankion.aiRelated wiki articles

• REST API index
• Authentication via Sanctum tokens
• Master Agent / Agentic Chat — multi-step reasoning via POST /v1/agentic/chat