AI that runs your whole backend
Build your backend in a conversation, from Claude Code, Cursor or the AI chat in the Anythink dashboard. The same MCP tools reach the whole CLI, from entities and fields to workflows, roles, secrets and payments.
The same task, three ways
Create a support_tickets entity and a workflow that emails the team when a ticket arrives.
The Anythink dashboard
Point and click. Good for seeing exactly what was built.
- ›Data Model → Add Entity: support_tickets
- ›Add Field: status (select)
- ›Create Entity
- ›Workflows → New Workflow: New ticket alert
- ›Trigger: a support_tickets record is created
- ›Add step: Send an Email
- ✓every new ticket emails support@example.com
The Anythink CLI
Scriptable and repeatable. Good for CI and for promoting staging to production.
$ anythink entities create support_tickets --rls$ anythink fields add support_tickets status --type varchar \--display select --options new,open,closed$ anythink workflows create "New ticket alert" \--trigger Event --entity support_tickets \--event EntityCreated --enabled$ anythink workflows step-add 61 notify-team \--action SendAnEmail --start --enabled \--params '{"to":"support@example.com","template":"new-ticket"}'- ✓Entity support_tickets created, with row-level security
- ✓Field status added
- ✓Workflow New ticket alert created and enabled
- ✓Step notify-team added
Claude, through MCP
Describe what you want, in Claude Code or the dashboard's AI chat. It runs the same commands and tells you what it did.
Why the interface can be plain text
“When the point of contact between the product and the people becomes a point of friction, then the industrial designer has failed.”
A dashboard has to be learned. An API has to be read. A conversation doesn't: you say what you want, and the Anythink MCP server carries it out, as real entities, fields and workflows. You don't have to learn Anythink before you can use it; you can describe the outcome, in your own words, and get a working system back.
One backend, reachable from plain text
The dashboard is the simple interface. The API, CLI and MCP go further: text with an AI assistant is the interface, and the output is a working system.
A public MCP server, not a demo
anythink-mcp is a published tool that speaks stdio, so it works with Claude Code, Cursor and any other MCP client. It isn't a cut-down preview: it reaches the same commands you'd run yourself.