Natoify Documentation

Everything you need to go from a one-line prompt to a scheduled, multi-step automation that emails you the result.

Quickstart

1. Sign up at /login.
2. Open New task, describe what you want in plain English.
3. Natoify proposes a plan. Click Run it to execute.
4. Watch the live computer panel as the agent works.

Bring your own keys

Natoify never charges you a markup for third-party APIs. Add your keys once in Settings:

• Firecrawl — best for general web search and structured extraction.
• Scrape.do — bulletproof scraping behind tough anti-bots.
• Resend — transactional email delivery.
• OpenAI / Anthropic / Perplexity — optional, used when you'd rather route to your own LLM account.

Keys are encrypted at rest with pgcrypto and only decrypted inside the agent runner.

Multi-step workflows

A task can hold a JSON array of steps. The agent runs each step in sequence, passing the previous output as context. Use this for jobs like "research → draft → email" where you want explicit hand-offs.

Triggers

Three ways a task can fire:

• Manual — you click Run.
• Webhook — POST to the unique URL on the task. JSON body becomes context.
• Schedule — cron expression, evaluated every minute by our worker.

MCP server

Every workspace exposes a Model Context Protocol endpoint. Generate a personal API key under Settings → API keys, then point your MCP-aware client (Claude Desktop, Cursor, Cline) at:

https://YOUR_PROJECT.functions.supabase.co/mcp

Tools exposed: run_task, agent_oneshot, get_run.

Connections

Add external MCP servers (Tasklet.ai, Linear, GitHub, etc.) under Connections. Natoify discovers their tool list on save and exposes them to your agent.

API reference

All Natoify HTTP endpoints live under /functions/v1/. Auth is either a Supabase user JWT or a personal nato_ API key as Bearer.