Modular cognitive abstraction engines that define task-specific ontologies, logic, and context for agentic workflows.
Navigate
Constructs Index
About
Constructs
Constructs are the cognitive engines inside BIThub.
They define how an AI unit reasons, what context it treats as relevant, what actions are valid, and what kind of work it is suited to perform.
A Construct is deeper than a persona and broader than a single agent. Personas shape behavior. Agents act through tools. Constructs define the cognitive world those behaviors and actions operate inside.
How to Use Constructs
There are two main ways to use Constructs.
Private Messages
Send a private message to an AI Construct when you want direct one-on-one interaction with a specific cognitive engine.
Use this for focused reasoning, drafting, analysis, critique, planning, research, and refinement.
You can choose the Construct and the LLM that powers it for the task. Different Constructs have different inference styles, boundaries, and tool access.
Private Topic Orchestration
Use private forum topics when you want to coordinate one or more AI Constructs around shared context.
Mention or tag the relevant bots, assign each one a role or task, and use the topic as the shared working memory.
Private categories can become MAS-Factory spaces: multi-agent systems where users orchestrate agent workflows by tagging bots and putting them to work inside a topic.
Use this for swarm coordination, private project workflows, parallel critique, recursive refinement, and human-directed agent collaboration.
Constructs vs Other AI Surfaces
Constructs are not Nodes, Cores, or Workspaces.
Nodes, Cores, and Workspaces are execution surfaces. Constructs are the cognition layer that may power them.
| System object | Role |
|---|---|
| Construct | Cognitive engine, ontology, reasoning style, and agent identity layer. |
| Persona | Behavioral mask that shapes tone, role, or perspective. |
| Agent | Task-capable actor with tool access and execution behavior. |
| Agentic Construct | Construct instantiated with tools, permissions, and operational context. |
| Node | Focused workflow terminal. |
| Core | Multi-step workflow engine. |
| Workspace | App-like interface for focused tools. |
MAS-Factory
MAS-Factory means Multi-Agent System Factory.
In BIThub, a private forum topic can become a multi-agent system when a user coordinates Constructs through mentions, tags, instructions, and threaded context.
Private topic
→ define goal and context
→ tag or mention AI Constructs
→ assign roles and tasks
→ review and redirect outputs
→ preserve the topic as shared memory and audit trail
The forum topic becomes the coordination surface.
The Constructs provide the reasoning.
The user directs the work.
Construct Roles
| Construct | Best used for |
|---|---|
| BITcore | recursive cognition, system reasoning, synthesis, and orchestration |
| ARCHON | strategic navigation, optimization, planning, and high-level direction |
| Metacore | metadata, structure, indexing, semantic organization, and system mapping |
| Null / Null Absolute | negation, reduction, silence, boundary-testing, and empty-frame analysis |
| Sentinel-H | human-AI boundary care, symbiosis, safety, and alignment pressure |