The building blocks
Start here. These four terms describe what you make and what happens when you press run.Workflow
A workflow is an editable graph that connects tools, data, and agents into a repeatable process. You build it on the visual canvas or generate it from a prompt with the AI Composer.
Run
A run is one execution of a workflow. Each run streams its progress live and ends in a final state you can inspect. You can run from chat, the API, or an SDK.
Node
A node is one step in a workflow. Every node writes its result back into the run so later steps can use it. ModuleX has nine node types.
Reference
A reference — written
{{node_id.field}} — pulls a value from an earlier node into a later one. References are how data flows between steps.The nine node types
A workflow is assembled from a fixed set of nine node types. You do not need to memorize them now — each links to its own reference page.LLM
Call a language model with a prompt and optional structured output.
Agent
Run an autonomous step that can call tools and loop until it is done.
Knowledge
Retrieve relevant context from a knowledge base inside a run.
Tool
Call one action from a connected integration.
Function
Run a built-in function such as an HTTP request or a webhook.
Conditional
Branch on an expression or an LLM decision, or loop over data.
Interrupt
Pause the run to ask a person a question, then resume with their answer.
Transformer
Reshape, map, and combine data between steps.
Guardrails
Validate content with schema, regex, or PII checks before it moves on.
The two AI helpers
ModuleX has two assistants, and they are easy to mix up. The short version: the AI Composer builds workflows; the Assistant gets work done without one.AI Composer
The text-to-workflow agent. Describe what you want in plain English and the AI Composer turns it into a complete, editable workflow graph — then keeps editing the canvas as you chat with it.
Assistant
A workflow-independent agentic chat. The Assistant searches your connected tools, decides the next step, calls tools, drafts outputs, and pauses for your approval before sensitive actions — no workflow required.
When do I use which one?
When do I use which one?
Use the AI Composer when you want a reusable process you can run again and again — for example, “every time a form is submitted, summarize it and post to Slack.” The Composer produces a workflow you own and can edit on the canvas.Use the Assistant for one-off or open-ended tasks where you do not want to build anything first — for example, “find the three most recent issues in this repo and draft a reply.” The Assistant acts directly using your connected tools.Both can stop and ask you for input or approval mid-task. That pause-and-resume behavior is called human-in-the-loop.
Knowledge and RAG
Knowledge is your own content — documents, files, and data — made searchable so a workflow or the Assistant can pull in relevant context automatically. The retrieval technique behind this is RAG (retrieval-augmented generation).Knowledge base
A knowledge base is the unit RAG searches over. You add documents to it; ModuleX splits them into chunks and indexes them for retrieval.
Managed vs your own store
Use the ModuleX-managed store (
modulexdb), or bring your own vector store such as Qdrant, Pinecone, MongoDB Atlas, or Weaviate. Managed retrieval and ingest are billed in credits; bring-your-own stores are not.Connecting the outside world
A workflow or the Assistant is only as useful as what it can reach. That is what integrations are for.Integration
An integration is a connector to an external service — Slack, GitHub, Google, and more. ModuleX ships 175 integrations.
Tool
A tool is one callable action an integration exposes (for example, “create an issue”). A single integration usually exposes many tools.
Credential
A credential is the stored, encrypted connection that lets ModuleX act on your behalf in an external service — for example an OAuth2 connection or an API key.
BYOK
BYOK (bring your own key) lets you connect your own model or service accounts. Usage you bring is billed by that provider, not metered as ModuleX credits.
How usage is paid for
Managed usage runs on credits. Bring-your-own-key usage does not consume credits — it is billed directly by your provider.Credit
A credit is the unit that meters ModuleX-managed usage. **100 credits equal 0.01). Each run, managed model call, and managed retrieval consumes credits.
Plan
Your plan sets your monthly credit allowance and limits. The plans are Free, Pro, Max, and Enterprise.
Wallet
The prepaid wallet covers usage beyond your plan’s allowance once you turn on extra usage. Top it up manually or automatically.
The usage gate
Before a run, Composer turn, Assistant turn, or managed knowledge call starts, ModuleX checks that you have credit and capacity. If you do not, the request is declined with a clear billing error instead of running and failing.
The usage gate is live on the run, AI Composer, Assistant, and managed-knowledge surfaces. When it declines a request you get a
402, 403, or 429 response carrying a structured reason. Plain create-read-update-delete and organization-settings calls are not gated this way. See Usage gating & limits and Errors & status codes for the exact shapes.Where everything lives: organizations
Everything in ModuleX belongs to an organization. It is the tenant and billing boundary — your workflows, credentials, knowledge bases, plan, wallet, and credits are all scoped to one organization.Organization
An organization is the shared workspace and billing unit. People you invite become members of it, and every run is metered against its plan and credits.
Roles
Each member has a role. The live roles are owner and admin. Actions such as using the AI Composer, the Assistant, and schedules require owner or admin.
Earlier ModuleX had a separate “member” role. It has been retired — the current roles are owner and admin only. If you see “member” referenced in older material, treat it as out of date.
Working together in real time
ModuleX is multi-user. Two things stream live, and they travel over different channels — worth knowing because the rest of the docs keep them separate.Run streaming
While a run executes, its progress streams to you event by event, so you can watch each node complete in real time.
Canvas collaboration
Several people can edit the same workflow canvas at once, with live cursors, presence, and locks so edits do not collide.
Keep going
How ModuleX works
See these concepts as one end-to-end flow, from a prompt to a running, observable workflow.
Quickstart
Create an organization, get an API key, and make your first authenticated call.
Glossary
The full canonical term list, including the exact spellings the docs standardize on.