Analytics preference

We only load Google Analytics after you opt in. You can change this choice at any time from Cookie settings.

Heym Features

Explore Heym's visual editor, AI Assistant, multi-agent orchestration, dashboards, RAG, execution traces, portal chat, evaluations, and workflow analysis.

Purpose-Built for AI Workflows

Not just automation tools with AI added on top. Heym is built from the ground up with AI as the execution model.

AI Assistant

Generate workflows from natural language or voice. The assistant streams nodes directly onto the canvas and works with any LLM provider.

Describe what you want in natural language or voice — the assistant generates nodes and edges and applies them to the canvas instantly. The AI streams its response and any valid workflow JSON is automatically parsed and applied, so you go from idea to working automation in seconds.

  • Natural language workflow generation from plain text or voice
  • Streams response and auto-applies generated nodes to the canvas
  • Works alongside manual editing for hybrid building
  • Filters attached Agent skills down to SKILL.md context when editing complex workflows
  • Supports any configured LLM credential and model
Read the documentation

Visual Workflow Editor

Drag-and-drop canvas with seven categories of built-in nodes, isolated output testing, sub-workflows, and an expression DSL.

Drag-and-drop canvas powered by Vue Flow with a broad library of built-in node types across 7 categories. Build complex workflows without writing code, pin node outputs for isolated testing, and track every change with built-in version history.

  • A broad node library spanning triggers, AI, logic, data, integrations, automation, and utilities
  • Expression DSL for dynamic data transformation between nodes
  • Data Pin lets you freeze a node output and test downstream logic without re-running
  • Extract to Sub-Workflow turns any selection into a reusable workflow
  • Full keyboard shortcuts: run, save, undo, copy, paste, and inline node search
Read the documentation

User-Built Dashboards

Grafana-style dashboards where each widget runs a Heym workflow, with a 12-column grid, 10+ chart types, and AI-generated widgets.

Build Grafana-style dashboards where every widget is powered by its own hidden Heym workflow. Add charts manually or describe the metric in plain language and let AI generate the widget workflow for you.

  • Drag and resize chart widgets on a 12-column dashboard grid
  • Render bar, line, area, pie, table, numeric, gauge, scatter, proportion, and bar gauge widgets
  • Each widget ends in a Chart Output node, so any workflow result can become a chart
  • Server-side TTL caching keeps dashboard loads responsive
  • AI generation and fine-tuning turn plain-language metric requests into widgets
Read the documentation

Agentic Kanban Board

Cards are persistent agentic jobs: moving one triggers the column workflow chain and writes results back automatically.

Cards are persistent agentic jobs — moving a card into a column runs that column's ordered workflow chain with the card's full context, and results are written back to the card.

  • Configure any column with an ordered list of workflows that run sequentially on card entry
  • Each workflow's output feeds the next; a failure stops the chain and turns the card red
  • Cards carry content, comments, activity history, and previous outputs into every run
  • Follow-up rounds re-run the current column chain with everything accumulated so far
  • Card status shows at a glance: green success, red failure, amber running or paused for review
Read the documentation

Multi-Agent Orchestration

One orchestrator delegates visually to named sub-agents and sub-workflows with skills, MCP connections, and automatic context compression.

One orchestrator agent delegates tasks to named sub-agents and sub-workflows, all wired visually on the canvas. Each agent can use Python tools, connect to MCP servers, load skills, and call other workflows — up to 5 levels of nesting depth.

  • Named sub-agents with independent tool calling and reasoning
  • Sub-workflows let agents invoke entire workflows as tools
  • Skills provide portable instruction files and Python tools via drag-and-drop
  • MCP client connections give agents access to external tool servers
  • Configurable reasoning effort and automatic model fallback
  • Automatic context compression keeps long-running agents within model limits
Read the documentation

Built-In RAG Pipeline

Upload documents to managed Qdrant or pgvector stores, then use semantic search, metadata filters, and optional Cohere reranking.

Upload PDFs, Markdown, CSV, or JSON to managed vector stores directly from the dashboard — backed by Qdrant or by Postgres (pgvector) inside Heym itself, no external service required. Wire a RAG node into any workflow to perform semantic search with metadata filters and optional Cohere reranking, then feed the results into LLM or Agent nodes.

  • Qdrant or built-in Postgres (pgvector) vector store management with one-click document upload
  • Semantic search returns text, relevance score, and metadata per result
  • Optional Cohere reranking for higher-precision retrieval
  • Multi-format document support: PDF, TXT, Markdown, CSV, and JSON
  • Share vector stores with team members for collaborative knowledge bases
Read the documentation

Persistent Agent Memory

Graph-based memory automatically extracts entities and relationships, with isolated scopes, a visual graph editor, and async execution.

Enable graph-based memory on any agent node to automatically extract and store entities and relationships from conversations. The knowledge graph persists across workflows and can be viewed, edited, and managed manually in the visual editor.

  • LLM-powered entity extraction with automatic relationship detection
  • Semi-aggressive deduplication merges similar entities automatically
  • Per-agent memory scope with isolated knowledge graphs
  • Visual graph editor for manual node and edge management
  • Async execution ensures workflows never slow down
  • View memory directly from canvas with a single click
Read the documentation

LLM Batch API Mode

Send prompt arrays through the OpenAI Batch API, with dedicated status outputs for live progress on high-volume, lower-cost work.

Switch the LLM node to OpenAI’s Batch API: one array of prompts per run, a main path for merged results, and a dedicated status branch for live progress while the batch job runs—built for high-volume, cost-efficient workflows.

  • Enable batch mode on supported models with your OpenAI API credential and Heym will guide availability in the node panel
  • One run can take an array expression such as $input.items.map("item.text") as the user message
  • Dedicated STATUS output emits pending, processing, completed, and failed style updates for side workflows
  • Final payload includes per-item results, counts, and batch job metadata from OpenAI
  • Suited to lower-cost bulk work, live progress notifications, and status logging branches
Read the documentation

AI & Automation

Direct LLM control and agent reliability

Developer & Integration

External tools and runtime systems

Secure & Platforms

Enterprise readiness and user interfaces

What's in the box

AI workflows that reason and run
Visible end to end

Animated Heym full feature set showcase
AI ASSISTANT

Describe it. Heym builds it.

Type or speak your automation in plain English — the Assistant generates the entire canvas: nodes, edges, credentials, expressions. Modify existing flows the same way: "add a HITL checkpoint before Output."

  • Works with any LLM credential you configure
  • Streams responses · auto-applies valid JSON
  • Voice input, transcribed in-browser
MULTI-AGENT

Agents that delegate to agents.

One orchestrator, five levels of nested sub-agents and sub-workflows. Each agent has its own model, tools, and Python execution context. Handoffs are visual — you can see the control flow, not infer it from logs.

  • Python tool calling per agent
  • MCP tool servers as drop-in tools
  • Auto context compression at 80% window
RAG · NATIVE

Vector stores, not another tool tab.

Upload PDFs, Markdown, CSVs directly. Managed vector stores — backed by Qdrant or by Postgres (pgvector) inside Heym — live alongside your workflows. The RAG node wires retrieval into any flow with metadata filters and optional Cohere reranking.

  • One-click upload · chunking & embeddings handled
  • Hybrid search (dense + BM25 + rerank)
  • Team-shared knowledge bases
MCP · FIRST-CLASS

Your workflows become AI tools.

Every workflow is exposable over MCP. Paste the SSE URL into Claude Desktop, Cursor, or any MCP client. Your AI can now run your automations from its own interface. Agents can also consume any external MCP server as tools.

  • Bidirectional: MCP server + client
  • Per-workflow API keys
  • No SDK, no adapter code
HUMAN-IN-THE-LOOP

Approval checkpoints as a primitive.

Drop a HITL node and execution pauses, generates a public review link, sends it via your configured channel. A reviewer can accept, edit, or refuse. AI output never ships without someone accountable signing off.

  • Public shareable review link
  • Accept · edit · refuse · timeout
  • Reviewer history in the trace
OBSERVABILITY

851 traces on average per real alpha user.

Every LLM call, every agent step, every token and millisecond — captured automatically. Filter by source, credential, model. The trace viewer shows the full prompt, response, tool calls, and timing waterfall. Replay any run against a different model.

  • Span timeline per node execution
  • Data Pin · freeze outputs for testing
  • Built-in evals with LLM-as-Judge

Everything in One Dashboard

One control center for everything you build. Workflows, credentials, vector stores, chart dashboards, alerts, teams, analytics, and evals all sit behind a single navigation bar of 18 tabs.

Each tab does one job. Sort workflows into folders, turn their results into live charts, get told when errors or costs cross a threshold, manage who on your team sees what, and score your prompts against several models. No external tools to bolt on.

Platform guides on GitHub
Click any card to view →

Feature guides on GitHub

Each card opens the canonical Markdown guide in the heymrun repository — AI assistant, agents, dashboards, canvas, guardrails, portal, security, and more.

Reference

40 Topics

Advanced reference documentation covering the AI assistant, agent architecture, canvas features, guardrails, parallel execution engine, expression DSL, credential sharing, portal chat UI, skills system, keyboard shortcuts, and security hardening.

01AI AssistantNatural language workflow builder inside the canvas editor. Describe what you want and the AI generates nodes and edges automatically. Supports voice input and streaming responses.02Agent ArchitectureDeep dive into agent orchestration patterns including lead/sub-agent delegation, tool calling, MCP integration, skills system, and configurable execution limits.03Canvas FeaturesVisual editor capabilities including drag-and-drop node placement, edge routing, zoom controls, minimap, snap-to-grid, node pinning, and output inspection panels.04GuardrailsContent safety filtering for LLM and Agent nodes. Blocks violence, hate speech, sexual content, prompt injection, and other unsafe categories with configurable error routing.05Parallel ExecutionDAG-based scheduling that automatically runs independent nodes concurrently in a thread pool. Downstream nodes start as soon as dependencies complete.06Human-in-the-LoopApproval checkpoints that pause workflow execution and generate a public review link. A human reviewer can approve or reject the pending action before it continues.07Expression DSLReference upstream data with $nodeLabel.field syntax. Supports string interpolation, ternary operators, array methods, object access, and built-in helper functions.08Credentials and SharingEncrypted credential storage with AES-256 Fernet. Share API keys with individual users or entire teams while keeping secret values masked and secure.09Portal Chat UITurn any workflow with an Input and Output node into a public chat interface. Embed on websites or share a link for end-user self-service powered by your workflows.10Drive & Generated FilesSkill-generated files in the Drive tab: search, download, share links with optional password and expiry. Use the Drive node in workflows to delete files or update access constraints programmatically.11Skills SystemPortable capability bundles consisting of SKILL.md instructions and optional Python tools. Drag and drop onto Agent nodes to extend context and toolbox.12Keyboard ShortcutsPower user shortcuts for node selection, deletion, duplication, undo/redo, zoom, panning, and canvas navigation to speed up workflow creation.13Edit HistoryTrack all changes to workflows with timestamps and user attribution. Restore previous versions and compare differences between workflow revisions.14Webhooks and TriggersConfigure JSON and SSE webhook endpoints, Telegram, Discord, and Slack webhooks, cron schedules, IMAP mailbox polling, outbound WebSocket client triggers, and RabbitMQ consumers for event-driven automation.15Execution TokensScoped JWTs that grant access to a single workflow's execute endpoint. Create short-lived or long-lived tokens directly from the cURL dialog, select one to embed it in the command, and revoke at any time without affecting your session.16SSE StreamingStream workflow execution over stream with execution_started, node_start, node_complete, and execution_complete events. The editor can generate ready-to-run cURL commands and per-node start messages.17SettingsPer-user settings dialog opened from the gear icon with four tabs: Profile and User Rules, Security and password changes, Voice for ElevenLabs TTS and STT, and Observability for read-only OpenTelemetry status.18Environment VariablesComplete self-hosting configuration reference covering required keys, database settings, networking, storage, MCP limits, Python tool sandboxing, plugins, tracing, and frontend build-time variables.19OpenTelemetry TracingEmit a root span per workflow run and a child span per node over OTLP/HTTP to Jaeger, Tempo, Honeycomb, or Datadog. W3C trace-context propagation, no instrumentation code, disabled by default.20SecuritySecurity architecture overview covering JWT authentication, HttpOnly cookies, credential encryption, role-based access control, and deployment hardening recommendations.21Workflow OrganizationOrganize workflows with folders, sub-folders, tags, and pinned favorites. Search workflows with the command palette and filter by status or recent activity.

Page1 / 7

Explore all 21 capabilities (click the slide arrows or an index chip to reveal):

Start with the IntroductionDiscovery endpoints for agents

Prefer raw Markdown? Browse the docs source on GitHub.