FAQ

    Frequently Asked Questions

    Straight answers about how URBLD runs the business end-to-end.

    AI Agent Connections

    Connect ChatGPT, Claude, Cursor, and future AI agents directly to your business.

    URBLD ships a live agent connection endpoint so any AI assistant can securely read leads, look up jobs, and act on your operations — using your team's identity, permissions, and audit trail. Powered by the open Model Context Protocol (MCP).

    Works with ChatGPT, Claude, Cursor, and any MCP-compatible client.

    How does URBLD connect to ChatGPT, Claude, and Cursor?

    URBLD runs a live MCP (Model Context Protocol) server at https://gdupthkvjnqtlqimcgtg.supabase.co/functions/v1/mcp. Compatible AI clients discover it, self-register through OAuth 2.1 dynamic client registration, and let the user approve access on URBLD's consent screen. From that point on, the agent calls URBLD tools as the signed-in user, with tenant-isolated data access and full audit logging.

    Compatible with every major AI client

    Because URBLD speaks the open MCP standard, adding a new AI client is not an integration project — it's a login.

    ChatGPT

    Custom GPTs and ChatGPT with the MCP connector can attach URBLD as a tool source.

    Claude

    Claude Desktop and Claude for Work discover URBLD via dynamic client registration.

    Cursor

    Cursor's agent surface connects to URBLD's MCP endpoint over OAuth 2.1.

    Any MCP client

    Anything speaking the open Model Context Protocol works — no per-vendor code.

    How the connection works

    Four steps. No API keys copied around. No shared service accounts.

    1. 1. Add URBLD

      Paste the MCP URL into ChatGPT, Claude, or Cursor. The client discovers URBLD automatically.

    2. 2. Sign in

      Log in with your URBLD account. Dynamic client registration handles the developer setup for you.

    3. 3. Approve access

      URBLD's consent screen tells you what the agent can do. Approve or deny — you're in control.

    4. 4. Start using

      The AI can now call URBLD tools as you, respecting your role, permissions, and org boundary.

    What AI agents can do today

    Read-only tools ship now. Write tools — creating leads, scheduling jobs, sending contracts, recording payments — roll out with per-tool audit logging and OAuth scope gating.

    whoami

    Confirm the connection and see which organization the agent is acting inside.

    list_leads

    Read recent leads scoped to the signed-in user's organization.

    list_jobs

    Read recent jobs, respecting tenant RLS and role permissions.

    Built for auditability, not vibes

    • OAuth 2.1 with PKCE and dynamic client registration — no long-lived API keys copied between systems.
    • Tenant isolation. Every tool runs under the user's JWT and inherits row-level security.
    • Consent screen. Users see exactly which client is connecting and can revoke access at any time.
    • Audit trail. Every mutation an agent performs is recorded to the tenant audit log with the resolved identity.
    MCP endpoint
    https://gdupthkvjnqtlqimcgtg.supabase.co/functions/v1/mcp
    
    Auth:  OAuth 2.1 + PKCE
    DCR:   supported
    Issuer: https://gdupthkvjnqtlqimcgtg.supabase.co/auth/v1

    Frequently asked

    What is the Model Context Protocol (MCP)?

    MCP is an open standard, originally introduced by Anthropic and now adopted by OpenAI, Cursor, and others, that lets AI assistants securely discover and call tools exposed by external systems. URBLD implements MCP so any compatible AI client can act on your business data with the same authentication, permissions, and audit trail your team uses.

    Do I have to write any code to connect ChatGPT or Claude?

    No. URBLD supports OAuth 2.1 with dynamic client registration, so ChatGPT, Claude, and Cursor self-register the first time you connect. You sign in with your URBLD account, approve access on the consent screen, and the assistant can call URBLD tools immediately.

    Is my data isolated to my organization?

    Yes. Every MCP tool call runs under the signed-in user's identity and inherits the tenant's row-level security. An agent connected to your workspace can never read or write another workspace's data.

    Which AI clients work with URBLD's MCP server today?

    Any MCP-compatible client. That includes Claude Desktop, ChatGPT's MCP connector, Cursor, and other agent frameworks that speak the standard. As new clients ship MCP support, they work with URBLD out of the box.

    What can AI agents do inside URBLD?

    Today's read-only tools cover identity, leads, and jobs. Additional write tools — creating leads, updating customers, scheduling jobs, sending contracts, recording payments — are rolling out with per-tool audit logging and role-scoped OAuth permissions.

    Share this page