Connect AXOIX to Claude, ChatGPT & Cursor

What this is

Instead of copy-pasting data between AXOIX and your AI assistant, you can connect the assistant directly to your tenant. It authenticates as you, via OAuth 2.1, and can only call the specific tools covered by the scopes you approve at connection time. This is separate from BYOK (bring-your-own-key AI features inside your AXOIX dashboard) — see our Privacy Policy section 4 for how the two differ.

Connect from Claude

1. In Claude, go to Settings → Connectors → Add custom connector.
2. Enter the server URL: https://mcp.axoix.io/mcp
3. Click Connect. You’ll be redirected to sign in with your AXOIX tenant subdomain, email, and password.
4. Review and approve the scopes you want to grant, then approve the connection.
5. Claude can now call AXOIX tools in conversation — you’ll see a confirmation prompt before any write or destructive action runs.

Connect from Cursor

Add the following to your Cursor MCP configuration:

{
  "mcpServers": {
    "axoix": {
      "url": "https://mcp.axoix.io/mcp"
    }
  }
}

Cursor will initiate the same OAuth 2.1 + PKCE flow on first use and prompt you to sign in and approve scopes.

Connect from ChatGPT or other MCP clients

Any client implementing the MCP Streamable HTTP transport with OAuth 2.1 (RFC 8414 / RFC 7591 dynamic client registration) can connect using the same server URL, https://mcp.axoix.io/mcp. Discovery metadata is published at https://mcp.axoix.io/.well-known/oauth-authorization-server.

Scopes

What’s available

• Identity & platform — whoami, list_modules, and the search/describe/invoke meta-tools that reach the full curated operation catalog.
• CRM — Search customers, look up contact and interaction history.
• Hotel operations — Check room availability, list and manage bookings.
• Finance — Sales summaries, outstanding receivables, invoice lookups.
• CHORUS — Generate marketing content, review scheduled posts.
• Remedy (support) — Search and respond to support tickets.
• Platform graph — Explore how your enabled modules and data relate to each other.

Beyond the tools listed above, a search/describe/invoke meta-tool group gives the assistant access to a broader curated catalog of read and write operations, each individually allow-listed and scope-gated — the assistant never gets unrestricted API access.

Security

• OAuth 2.1 with mandatory PKCE (S256) — no plaintext or “plain” code challenges accepted.
• Every access token is bound to your tenant, its issuing audience, and an expiry; tokens for a different tenant or a different resource are rejected outright.
• Destructive and financial operations require an explicit confirmation step before they execute.
• Every tool call is written to your tenant’s audit log, the same trail your dashboard actions produce.
• You can revoke access at any time from your AI assistant client’s connector settings.

Support

Questions or issues connecting? Email [email protected].