Can AI agents automate Retool?

Retool is automatable today through a first-party tenant MCP server and a REST organization API. This record is Retool Cloud and self-hosted Retool 4.0.0+ as exposed by https://<tenant>.retool.com/mcp for app building and organization and user management. It is not Workato, Appsmith, or Power Apps, and it is not Retool's separate MCP resource that only consumes external MCP servers. The first-party server uses Streamable HTTP and OAuth 2.0. Documented tools list, inspect, and grant access to apps; create and update resources; execute TypeScript against resources; create or continue React-app threads from a prompt; import React apps; publish threads; and, with the mcp:admin scope, list users, groups, workflows, folders, environments, audit logs, and invites. Scopes cannot exceed the signed-in account. Only organization administrators can select mcp:admin. Self-hosted MCP requires version 4.0.0 or later and the same prerequisites as the new app building experience. Classic drag-and-drop builder work is out of this MCP path. The Retool API at https://{instance}/api/v2 authenticates with a Bearer access token created by an organization admin under Settings → Retool API. It lists, clones, and deletes apps; manages users, groups, folders, spaces, resources, permissions, roles, and workflows; and publishes an OpenAPI spec. Cloud organizations use https://api.retool.com/api/v2/spec. The API is documented for cloud and self-hosted 3.18+. It does not replace the React app-building MCP tools. Retool's pricing page lists full platform APIs including write access on Enterprise. Official retool-cli installs from npm and can list, create, export, and delete apps after interactive login. Workflows accept inbound JSON webhooks authenticated with X-Workflow-Api-Key. iPaaS catalogs were not opened. Computer-use viability is unassessed. MCP scores 8 and is the best path for agents that should build or inspect Retool apps. Use the REST API for deterministic org provisioning.

Best path today: mcp · Overall automatability: 8/10