Aigur.dev

Provides a canvas-based interface where users drag AI operation nodes (LLM calls, data transformations, conditionals, loops) and connect them via edges to define execution flow. The builder likely uses a graph-based data model (DAG) to represent workflows, with real-time validation of node connections and type compatibility. Workflows are stored as JSON/YAML configurations that can be versioned and deployed without code generation.

Enables multiple team members to edit the same workflow concurrently using operational transformation or CRDT-based conflict resolution. The platform tracks cursor positions, node selections, and edits in real-time, showing which team member is working on which part of the workflow. Changes are synchronized across all connected clients without requiring manual merges or version conflict resolution.

Provides pre-built connector nodes for popular services (Slack, Google Sheets, Salesforce, HubSpot, etc.) that handle authentication, request formatting, and response parsing. Users select a connector, authenticate with the service, and configure the operation (e.g., 'send Slack message', 'append row to Google Sheet'). The platform manages API credentials securely and abstracts away service-specific API details.

Allows workflows to be executed on a schedule (daily, weekly, monthly, or custom cron expressions) without manual triggering. Users configure the schedule in the workflow settings, and the platform's scheduler triggers executions at the specified times. Scheduled executions are treated like any other execution, with full logging and monitoring available.

Organizes workflows, templates, and team members into workspaces with role-based permissions. Workspace admins can invite team members, assign roles (admin, editor, viewer, executor), and control access to workflows and resources. The platform enforces permissions at the workflow level, preventing unauthorized users from viewing, editing, or executing workflows.

Provides a standardized node type for LLM calls that abstracts away provider-specific APIs (OpenAI, Anthropic, Cohere, local models). Users configure the node with a prompt template (supporting variable interpolation from upstream nodes), model selection, temperature, max tokens, and other hyperparameters. The platform handles authentication, request formatting, and response parsing transparently, allowing non-technical users to chain LLM calls without managing API keys or request/response schemas.

Provides pre-built node types for common data operations: JSON path extraction, field mapping, filtering, aggregation, and format conversion (CSV to JSON, etc.). Users define transformations declaratively (e.g., 'extract field X from input, rename to Y, filter where Z > 10') without writing code. The platform likely uses a schema-based approach where users specify input/output shapes, enabling type checking and validation across the workflow.

Allows workflows to include decision points (if/else based on upstream data), loops (iterate over arrays with per-item processing), and error handling branches. Users define conditions using a visual rule builder (e.g., 'if field X equals Y, go to node A, else go to node B'). The platform executes branches conditionally and manages loop state, enabling complex multi-path workflows without explicit code.

Converts visual workflows into executable configurations that can be deployed to a serverless execution environment. The platform manages versioning (each save creates a snapshot), allowing users to roll back to previous versions or run multiple versions in parallel. Deployments are triggered manually or via webhooks/API calls, with execution tracked and logged for debugging and audit purposes.

Allows workflows to be triggered by incoming HTTP requests (webhooks) or called directly via REST API. Users configure the trigger endpoint, specify expected input parameters, and map incoming request data to workflow variables. The platform generates a unique webhook URL for each workflow, handles authentication (API keys or basic auth), and routes requests to the appropriate workflow version.

Provides a repository of pre-built workflow templates for common use cases (customer support, content generation, data processing, etc.). Users can browse templates, fork them into their workspace, and customize for their needs. The platform implements role-based access control (RBAC) for sharing workflows within teams, with granular permissions (view, edit, execute, delete). Templates can be marked as public (organization-wide) or private (individual).

Tracks workflow executions in real-time, displaying execution status (running, completed, failed), step-by-step logs with timestamps, and performance metrics (execution time per node, total duration, token usage for LLM calls). Users can drill into individual executions to debug failures, inspect intermediate data, and understand bottlenecks. Logs are searchable and filterable by status, date range, or node.

Allows advanced users to write custom JavaScript or Python code within a workflow node, executed in a sandboxed environment with access to upstream workflow data. The code node provides a standard interface for input/output, with libraries available for common tasks (HTTP requests, JSON processing, etc.). The sandbox isolates custom code from the platform and other workflows, preventing malicious or buggy code from affecting system stability.