Magic Loops

Converts plain English descriptions of repetitive tasks into executable automation workflows without requiring code. Uses LLM-based intent parsing to translate user descriptions into structured workflow definitions, then maps those definitions to pre-built action nodes (HTTP requests, data transformations, conditional logic). The system maintains a library of common automation patterns and learns from user corrections to improve future parsing accuracy.

Provides pre-built connectors to 100+ SaaS applications (Slack, Gmail, Notion, Airtable, etc.) with OAuth-based credential handling that abstracts away API authentication complexity. Each connector exposes a standardized action interface (trigger, filter, transform, send) that maps to the underlying app's REST API, with automatic request/response transformation and error handling. Credentials are encrypted and stored securely, allowing users to reference integrations by name rather than managing tokens.

Allows users to make arbitrary HTTP requests to any API endpoint (not just pre-built connectors) by specifying method (GET/POST/PUT/DELETE), URL, headers, and body. Supports templating in all fields using the same expression language as data transformation, enabling dynamic URL construction and request body generation based on previous step outputs. Handles common authentication patterns (API key, Bearer token, Basic auth) and automatically manages request/response encoding.

Executes workflows on two execution models: time-based scheduling (cron-like intervals: hourly, daily, weekly) and event-based triggering (webhook listeners that fire on external events). The system maintains a distributed task queue that dequeues scheduled jobs at specified times and maintains persistent webhook endpoints that capture incoming events and trigger corresponding workflows. Execution state is tracked per workflow run, enabling retry logic and failure notifications.

Provides a canvas-based interface where users drag pre-built action nodes (HTTP request, data filter, conditional branch, loop, etc.) onto a workflow graph and connect them with edges to define execution flow. Each node exposes configurable parameters (URL, headers, body template, condition logic) through a side panel. The editor validates the workflow graph for structural correctness (no orphaned nodes, valid connections) and provides real-time syntax checking for expressions and templates.

Provides a templating and expression language (likely Handlebars or similar) that allows users to map outputs from one workflow step as inputs to the next step. Supports field extraction from JSON responses, string interpolation, conditional value selection, and basic arithmetic operations. The system maintains a context object containing all previous step outputs, making them available for reference in downstream steps via dot notation or bracket syntax.

Allows users to execute a workflow with test data before scheduling or deploying it to production. The dry-run mode simulates each step without making actual API calls to external services (or makes calls to test endpoints if available), capturing the execution path and output at each node. Users can inspect intermediate results, validate that data transformations are correct, and identify logic errors before the workflow runs on real data.

Allows users to configure retry behavior for individual workflow steps or entire workflows when failures occur. Supports exponential backoff (delay increases with each retry), maximum retry counts, and conditional retry logic (retry only on specific error types). Failed workflows can be configured to send notifications (email, Slack) or trigger alternative workflows, enabling graceful degradation and alerting.

Maintains a complete log of every workflow execution including timestamp, trigger type (scheduled/webhook), execution duration, success/failure status, and full step-by-step execution trace. Users can filter execution history by date range, status, or trigger type, and drill into individual executions to inspect intermediate results and error messages. Audit logs are immutable and retained for compliance purposes.

Allows users to share workflows with team members with granular permissions (view-only, edit, execute, delete). Shared workflows maintain a single source of truth — changes made by one user are immediately visible to others. Role-based access control (RBAC) restricts who can modify, execute, or delete workflows. Audit logs track which user made which changes and when.

Provides a library of pre-built workflow templates for common automation scenarios (e.g., 'send Slack notification on new email', 'backup Airtable to Google Drive', 'sync Notion database to Slack'). Templates are parameterized — users customize them by providing app credentials and configuration values without modifying the underlying workflow logic. Templates can be created by users and shared with the community or within an organization.