Cloud/Create Connector/Manual Setup

Manual Setup

Ask AI about Vinkius

The fourth path in Create Connector: no API specification? No problem. Add your endpoints manually and connect your API to AI agents step by step.

Manual Setup is the fourth path inside Create Connector, and it is the fallback that keeps the door open. The console's card states it plainly: "No API specification? No problem. Add your endpoints manually and connect your API to AI agents step by step." It carries no tags because it makes no promises about speed: it is the honest path for the API that has no OpenAPI document yet.

R

Deploy Connector

Connect your API to AI agents in under 30 seconds.

Manual Setup, live. No spec? Name the connector, point at your API base URL with HTTPS validation, choose the auth type and pick your governance switches before the deploy runs its phases.

The mockup is the real form. Clear the base URL or type one without HTTPS to see the validation refuse it, then walk Security and Governance and deploy.

Your app, your way

The education panel frames it: "Name your connector and point us to your API. Add endpoints one by one. AI agents discover each tool as you build." The step asks for exactly two fields:

  • Server Name, with the note that "This name is what AI agents see when they discover your tools";
  • Base URL, required, and validated for real: anything without HTTPS gets "Enter a valid HTTPS URL (e.g. https://api.example.com)", an empty field gets "Base URL is required."

That is the whole first step. No spec parsing, no magic: you are introducing your API by name.

The same security, the same governance

From here, Manual Setup walks the identical steps as the OpenAPI path. Security asks for the Auth Type: None · Public API, Bearer Token, Basic Auth or Custom Header, with the standing promise "AES-256 encrypted at rest. Credentials never leave your vault." Governance offers the two default switches: Zero-Trust PII Redaction and Smart Array Truncation (Cost Control). Later you can tune both per connector in DLP Protection and FinOps Guard.

The deploy

Deploy to Edge runs the manual phases: provisioning the server, configuring the security layer, activating data protection, going live on Edge. The success screen is the standard one: your connector is Live!, the one-time Connection Token with the "Save this token now. It won't be shown again." warning, and the mcpServers config JSON ready to copy or download for Claude, Cursor, VS Code, Antigravity or any other client.

Growing the connector after the deploy

Manual Setup is the beginning, not the whole story. Your API is connected, and the tools your agents can call are managed from the connector itself: the AI Capabilities tab is where the connector's tool set lives and gets switched, and the Playground is where you prove each capability works before your AI leans on it. The connector appears in your Library like every other one, with Dashboard, Credentials and Connection Tokens ready.

Why this path exists

Not every API has a spec, and waiting for one would mean waiting to be AI-ready. Manual Setup keeps the Create Connector door open for the internal tool, the legacy service, the API whose documentation lives in someone's head. You connect it now, and the catalog grows by one more thing agents can actually use.

What comes next

This was the last of the four Create Connector paths. After the paths, the console's Seller Account covers listing your connectors on the marketplace and earning from every request.