MCP server

The Xizor MCP server puts your entire subscription business inside your AI agent. Connect Claude, ChatGPT, Gemini — any Model Context Protocol client — and your agent can read every metric and write: draft paywalls, translate them into any language, and publish them live.

Connect URL:  https://mcp.xizor.dev/mcp
Transport:    Streamable HTTP (MCP spec 2025-11-25)
Auth:         OAuth 2.1 — sign in with your Xizor account, no API keys

What the server can do

29 tools across analytics (read) and paywall management (write):

Apps & analytics (read) — list_apps, get_overview_stats, get_stats_timeseries (revenue, MRR, active subs, new subs, churn), get_trial_conversion, get_refunds, get_ltv, get_arpu, get_cohort_retention, get_paywall_funnel, get_breakdown (by country/product).

Paywalls & placements (read + write) — list_paywalls, get_paywall, create_paywall, duplicate_paywall (copy a paywall into a fresh draft — the fast path to a near-identical variant, cross-app too), update_paywall_draft, publish_paywall, get_paywall_texts, localize_paywall, list_placements, create_placement, assign_placement, set_placement_rules (context-based routing: "referred users get the referral paywall"), validate_paywall_document, get_paywall_schema_docs.

Catalog & subscribers (read) — list_products, list_entitlements, get_subscriber, search_transactions, list_events.

Writes are safe by design: paywall edits go to the draft; publish_paywall is an explicit, separate step that validates the document first. Published changes are live in your app within seconds (how save-to-live works).

How authentication works (OAuth, no API keys)

You never copy an API key. The first time your agent calls the server:

1. Your MCP client opens a browser window to the Xizor login.
2. You sign in with your normal Xizor account.
3. A consent screen shows the requested scopes — Read (xizor:read: apps, analytics, paywalls, subscribers, events) and Write (xizor:write: edit drafts, publish paywalls, manage placements and localizations). Grant read-only if you prefer.
4. The client receives scoped tokens; you can revoke the connection anytime from your Xizor account.

That's OAuth 2.1 with PKCE and dynamic client registration — every mainstream MCP client handles it automatically.

Connect: Claude.ai

1. Open Settings → Connectors → Add custom connector.
2. Paste https://mcp.xizor.dev/mcp.
3. Sign in with your Xizor account when prompted and choose your scopes.

Works on web and in the Claude desktop/mobile apps; the connector is then available in every chat (enable it via the tools menu).

Connect: Claude Code

claude mcp add --transport http xizor https://mcp.xizor.dev/mcp

Then run /mcp inside Claude Code to complete the OAuth sign-in in your browser. Done — claude "what's my MRR?" now works.

Connect: ChatGPT

1. Enable Developer Mode: Settings → Apps & Connectors → Advanced settings → Developer mode.
2. Go to Settings → Connectors → Add connector (Create in Developer Mode).
3. Set the server URL to https://mcp.xizor.dev/mcp, authentication: OAuth.
4. Complete the sign-in flow.

Connect: Gemini CLI

gemini mcp add --transport http xizor https://mcp.xizor.dev/mcp

Or add it manually to ~/.gemini/settings.json:

{
  "mcpServers": {
    "xizor": {
      "httpUrl": "https://mcp.xizor.dev/mcp"
    }
  }
}

The CLI opens the browser for OAuth on first use (/mcp auth xizor if you need to re-trigger it).

Example prompts

Analytics:

• "What's my MRR trend over the last 90 days, and which product drives it?"
• "Compare trial conversion this month vs. last month. Any country standing out?"
• "Which paywall version converts best on the onboarding placement?"

Paywall work (write scope):

• "Translate my onboarding paywall to Japanese and publish it."
• "Create a winter-sale paywall from the cinematic template with 40% off messaging, assign it to the promo placement, but don't publish yet — show me the draft first."
• "The CTA on the upsell paywall says 'Subscribe'. Change it to 'Start my free week' in all locales and publish."

Combined:

• "Look at the funnel for the paywall on settings. If the start rate is under 5%, rewrite the headline to focus on the free trial and publish a new version."

Agents always operate on drafts first and need an explicit publish — you can also just keep write scope off and use Xizor MCP as a pure analytics companion.

Tips

• get_paywall_schema_docs gives the agent the full document schema so it writes valid paywalls on the first try; validate_paywall_document lets it check before publishing.
• Coding agent integrating Xizor into an app? Point it at For coding agents and /llms-full.txt.