Core Concepts
The Dynamiq glossary: organizations, projects, workflows, apps, deployments, traces, access keys, and how they fit together.
Dynamiq has a small, consistent vocabulary. Learn these terms once and every page in the docs — and every label in the UI — will read the same way.
The hierarchy
Resources are organized in three levels:
Organization
└── Project
├── Workflows (and their versions)
├── Apps (deployed workflows) → Deployments, Runs, Traces, Sessions, Triggers
├── Knowledge Bases
├── Connections
├── Prompts
└── Skills- Organization — the top-level account scope. Members, roles, billing, and Access Keys live here. You can belong to several organizations.
- Project — a workspace inside an organization. Everything you build — workflows, apps, knowledge bases, connections, prompts, skills — belongs to exactly one project. The project selector at the top of the sidebar switches between them.
See Organizations & Projects for membership and role management.
Building
| Term | Meaning |
|---|---|
| Workflow | The buildable DAG you edit on the canvas. Listed under Agents in the sidebar. Not a "flow" or "pipeline". |
| Node | One step in a workflow: an LLM, an Agent, a tool, a retriever, a validator, a transformer. Each node has a configuration panel and typed inputs/outputs. |
| Connection | Stored credentials and configuration for an external service (an OpenAI API key, a Postgres URL, a Tavily key). Nodes reference Connections instead of embedding secrets. Managed under Connections. |
| Agent node | The ReAct agent node: an LLM that reasons in a loop and calls the tools you attach to it. The center of most Dynamiq workflows. |
| Release / Version | A saved snapshot of a workflow. Every Save produces a version; deployments pin to a specific version, which is what makes rollback safe. |
| Prompt | A versioned, reusable prompt template managed under Prompts and referenced from LLM and Agent nodes. |
| Skill | A reusable instruction pack that changes how an agent approaches a class of task. Used in Chat and by agents; managed under Skills. |
| Knowledge Base | An ingested, chunked, and embedded document collection that agents and workflows search semantically. The vector store is its storage backend, not its name. |
Deploying and running
| Term | Meaning |
|---|---|
| App | A deployed workflow: a stable resource with its own HTTPS hostname, access control, monitoring, and history. Not a "deployment endpoint" or "service". |
| Deployment | The act (and record) of deploying a workflow version to an App. An App accumulates deployments over time; the HISTORY tab lists them. |
| Run | One execution of a deployed app — one invocation, one trigger firing, or one chat turn against the app. |
| Trace | The recorded execution tree of a run: every node, agent loop, tool call, and LLM request with timings and token usage. Browsable in the app's TRACES tab. |
| Session | A conversation thread against an app. Multi-turn chat with a deployed agent groups its runs under a session so memory and history hold together. |
| Trigger | A scheduled or event-based app invocation — run an app on a cron schedule or when an external event arrives, instead of by direct HTTP call. See Triggers. |
Chat-specific
| Term | Meaning |
|---|---|
| Chat | The super-agent surface at /chat. Two modes: Dynamiq Agent (the built-in agent on a model you pick) and Custom Agents (your deployed apps). |
| Connector | An OAuth app integration used in Chat — Google Drive, Gmail, Notion, Slack, databases. Distinct from a Connection: Connectors are per-user app links for Chat; Connections are project credentials for workflow nodes. |
Access Key vs. Personal Access Token
Both are bearer credentials; they answer different questions.
| Access Key | Personal Access Token (PAT) | |
|---|---|---|
| Belongs to | The organization (optionally scoped to one project) | You, the user |
| Used for | Calling deployed apps, inferences, knowledge bases, and services | Calling the management API on your behalf |
| Acts as | The organization — not any specific user | You, with your exact permissions, org membership, and roles |
| Typical use | Production integrations and backend services invoking a deployed resource | Scripts, CLI tools, and automation that manage projects, workflows, and datasets |
| Created in | Settings → Access Keys | Profile → Personal Access Tokens |
Rule of thumb: invoking a deployed resource → Access Key; managing the platform → Personal Access Token. Details in API Keys & Tokens.
How the pieces connect
A typical lifecycle, in vocabulary order: you build a Workflow out of Nodes in a Project, using Connections for credentials and maybe a Knowledge Base for grounding. Saving creates a Version. Deploying that version creates (or updates) an App — that act is a Deployment. Clients call the App with an Access Key; each call is a Run, recorded as a Trace, optionally grouped into a Session, and possibly initiated by a Trigger rather than a direct call.