Business AgentAI Office Assistant Platform

An agentic runtime that plans tasks, executes them via MCP tools, and persists execution state

PORA Loop

Plan-Observe-Reason-Act cycle for agentic execution. LLM selects tools, executes them, observes results, and decides next steps.

Switch execution contexts across local, SSH, container, teleport, and direct modes.

Schema-driven CRUD for enterprise entities - product, sales_order, invoice, expense, supplier, and more.

Six modules - sales, crm, procurement, inventory, finance, pm - with dynamic routing and multi-view support.

Orient-Gather-Consolidate-Prune memory index with VITAL protection for critical content.

Thinking clear → Tool result prune → Stale prune → LLM summarize for long conversations.