n8n

Provides a drag-and-drop canvas interface for constructing directed acyclic graphs (DAGs) of interconnected nodes, where each node represents an integration or transformation step. The frontend uses Vue.js state management to track node positions, connections, and parameter configurations in real-time, with the workflow definition serialized as JSON and persisted to the backend. Supports dynamic node type registration from the node registry, enabling users to discover and compose 400+ integrations without code.

Executes workflows using a pluggable execution model supporting multiple runtime modes: single-process (main thread), worker threads, and distributed execution across multiple instances. The core execution engine (packages/core) orchestrates node execution sequentially or in parallel based on workflow topology, managing data flow between nodes through an expression system that evaluates JavaScript-like syntax. Supports both synchronous and asynchronous node execution with built-in timeout handling, error recovery, and execution state persistence to the database for resumability.

Exposes HTTP webhooks for each workflow that accept incoming requests and trigger workflow execution with the request payload as input. Webhooks support request validation (signature verification, IP whitelisting), custom response mapping (transform workflow output into HTTP response), and rate limiting. The webhook system integrates with the execution engine to queue executions and return results synchronously or asynchronously based on workflow configuration.

Enables workflows to be triggered on a schedule using cron expressions (e.g., `0 9 * * MON-FRI` for weekday mornings) with timezone awareness for global teams. The scheduler runs as a background job that evaluates cron expressions and enqueues workflow executions at the appropriate times. Supports multiple schedules per workflow, execution history tracking, and manual trigger overrides for testing.

Provides a TypeScript SDK (@n8n/node-dev) for developing custom nodes that extend n8n's capabilities beyond the built-in integrations. Custom nodes are packaged as npm modules with metadata describing node properties, parameters, and credentials. The node registry dynamically loads custom nodes from installed npm packages, enabling community contributions and enterprise-specific integrations. Includes scaffolding tools, testing utilities, and documentation for node development.

Provides a key-value data store (Data Store module) that persists data across workflow executions, enabling workflows to maintain state between runs. Data store operations (get, set, append, delete) are exposed as nodes that can read and write arbitrary JSON data with optional TTL (time-to-live) for automatic expiration. The data store is backed by the database and supports querying by key prefix for bulk operations.

Enables workflows to be versioned and synchronized with Git repositories, allowing teams to manage workflow definitions as code. Workflows can be exported to JSON files and committed to Git, with automatic synchronization between n8n and the repository. Supports branching, merging, and rollback to previous workflow versions through Git history. Integrates with GitHub, GitLab, and Gitea for seamless source control workflows.

Provides a testing framework for validating workflows before deployment, including mock data generation, test execution, and assertion checking. Tests can be defined as JSON configurations that specify input data, expected outputs, and assertions (e.g., 'output should contain field X'). The framework supports running tests against workflow definitions without executing external integrations, enabling fast feedback loops during development.

Provides a custom expression language (packages/@n8n/expression-runtime) that evaluates JavaScript-like syntax to dynamically transform and bind data between workflow nodes. Expressions are evaluated in a sandboxed runtime with access to input data, workflow variables, and helper functions, enabling users to reference previous node outputs (e.g., `{{ $node.HTTPRequest.json.data }}`), perform conditional logic, and apply transformations without writing separate code. The expression system integrates with the parameter input editor, allowing expressions to be used in any node parameter field.

Integrates an AI-powered workflow builder (AI Workflow Builder subsystem) that accepts natural language descriptions and generates complete workflow definitions by calling an LLM (Claude, GPT-4, etc.) with a system prompt that understands n8n's node types and workflow structure. The builder parses the LLM response to extract node definitions, connections, and parameters, then validates the generated workflow against the node registry before presenting it to the user for editing. Supports iterative refinement where users can ask the AI to modify specific workflow sections.

Implements an autonomous agent system (Instance AI subsystem) that uses an LLM to make decisions about which workflow nodes (tools) to execute based on user intent and execution context. The agent maintains a conversation history, evaluates available tools (nodes) against the current goal, and iteratively executes tools and processes results until reaching a terminal state. Integrates with LangChain nodes for chain-of-thought reasoning and supports multi-turn interactions where the agent can ask clarifying questions or report progress back to the user.

Provides a dedicated node package (@n8n/nodes-langchain) that wraps LangChain components (LLMs, chains, memory, vector stores, tools) as n8n nodes, enabling users to build complex LLM applications within workflows. Nodes include LLM providers (OpenAI, Anthropic, Ollama), chain types (ReAct, MapReduce), memory backends (Redis, in-memory), and vector store connectors (Pinecone, Weaviate, Chroma). The integration handles credential management, prompt templating, and output parsing, abstracting LangChain's Python API into a visual node interface.

Implements a centralized credential system (Credentials API and Security subsystem) that stores encrypted credentials in the database and resolves them at runtime using a pluggable backend system. Supports multiple credential types (API keys, OAuth tokens, database connections, SSH keys) with type-specific validation and encryption. Integrates with external secret stores (AWS Secrets Manager, HashiCorp Vault, environment variables) via a dynamic credential loader, enabling credentials to be fetched from external systems without storing them in n8n's database.

Implements a project-based authorization model where workflows are organized into projects with role-based access control (RBAC) for users and teams. Each project has configurable roles (Admin, Editor, Viewer) with granular permissions for workflow execution, editing, and credential access. Supports project-level sharing where workflows inherit permissions from their parent project, and team-based access where multiple users can collaborate on workflows with conflict resolution for concurrent edits.

Enables horizontal scaling of workflow execution by distributing jobs across multiple worker instances using Redis as a message broker. When a workflow is triggered, the main instance enqueues the execution job to Redis, and available workers dequeue and execute the job independently. Workers report execution progress and results back to the main instance via Redis, enabling real-time monitoring of distributed executions. Supports worker auto-scaling based on queue depth and execution latency.

Maintains a complete execution history for every workflow run, storing execution metadata (start time, duration, status), node-level logs (input/output data, errors), and audit events (who triggered the workflow, when, from which IP). Execution logs are structured as JSON documents enabling full-text search and filtering by node, status, or time range. Supports configurable log retention policies and archival to external storage (S3, GCS) for long-term compliance.