Discover connectors

Name tools and schemas from the live catalog and Scalekit MCP. Then stop.

Guardrails

• MUST treat the live catalog and Scalekit MCP as the source of truth. Do not cache a connector list in the repo.
• MUST page list_tools / search_tools until the next-page token is empty. One call is not the full catalog.
• MUST NOT execute tools or authorize a connected account. Name integrate-agentkit for that.

Gotchas

• Read SDK credentials from SCALEKIT_ENVIRONMENT_URL, SCALEKIT_CLIENT_ID, and SCALEKIT_CLIENT_SECRET. Some samples use SCALEKIT_ENV_URL; use SCALEKIT_ENVIRONMENT_URL here.
• A connection is dashboard connector config. A connected account is one user authorized on that connection. This skill names tools. It does not create either.
• Use connector in prose. Use provider only when the SDK field is literally provider or providers.
• Filter.provider is the connector identifier from the catalog or search_connectors (for example GMAIL). It is not the dashboard Connection Name.
• Framework helpers such as actions.langchain.get_tools need a connected-account identifier and hide pagination. Use actions.tools.list_tools here.
• Output schemas are not in the tool definition. Use the connector's official API docs for response shape.
• Default language is Python. If the repo is Node, open references/node.md. If the language is unknown, stay on Python.