AI agents & MCP
Let your coding assistant spin up a realistic Postgres on demand: five MCP tools, referentially intact data, one prompt.
The problem
AI assistants scaffold schemas and app code fast, then hit a wall: nothing realistic to query. They either spin up an empty database or ask the developer for credentials and fixtures, which breaks the flow.
How Weavori handles it
- A built-in MCP server exposes introspect, estimate, generate, doctor, and sync to any MCP client (Claude Desktop, Cursor, VS Code, Qoder).
- Generation runs in FK dependency order, so agent-built apps query valid data immediately.
- Same semantics as the CLI, so an agent and a human get identical results.
Connect an assistant
Install the CLI
$npm install -g @weavori/cli
Start the MCP server from your assistant's config
$weavori mcp
Get started
i
Set up your assistant
The MCP setup guide walks through wiring Weavori into each supported client.