Getting started
Sign up once. Hand an API key to your agent. It handles everything else.
1. Sign up
Email OTP — no password. Takes ten seconds.
Head to /signup, enter your email and display name, and paste the six-digit code. You land in an empty workspace with no apps — that's by design. From here, your AI agent takes over.
2. Connect your AI agent
Claude.ai needs no API key — just add the connector and log in with OAuth.
Claude.ai (web)
Go to Settings → Connectors → Add custom connector. Enter name Rowly and URL:
https://rowly.ai/mcp
Log in with OAuth when prompted — no API key needed.
Claude Desktop & other MCP clients
Create a user-scoped API key first: go to Settings → API keys → New key and choose type rowly_live_u_…. Then add to your claude_desktop_config.json:
{
"mcpServers": {
"rowly": {
"url": "https://rowly.ai/mcp",
"headers": {
"Authorization": "Bearer rowly_live_u_<your-key>"
}
}
}
}
Agent Skill (optional)
Install the public Rowly skill so supported coding agents get Rowly-specific setup, schema, records, workflow, and MCP guidance. The public skill never includes API keys or tenant-private Skill records.
npx skills add https://rowly.ai --list
npx skills add https://rowly.ai --skill rowly -g -a codex -y
Verify the catalog at https://rowly.ai/.well-known/agent-skills/index.json.
Restart your client. You should see rowly in the tool list with the current MCP catalog attached.
3. Ask the agent to set up your workspace
One prompt. The agent creates the app, defines docs, and verifies schema — no UI needed.
Claude · mcp://rowly
→ rowly.create_app({ name: "My Business" })
created → app_7c2d
→ rowly.create_doc({ type: "VendorInvoice", fields: 5 })
5 fields defined, D1 table ready
4. Let the agent run your business
Ask in plain language — the agent reads invoices, creates records, and updates status.
From here, your agent handles day-to-day operations:
Add invoice INV-2025-042 from บริษัท ABC จำกัด, amount ฿15,000, date today.
Show all pending invoices.
Approve invoice INV-2025-042.
The admin UI at /data reflects every record in real time — watch what the agent writes, audit the history, or step in manually when needed.