Resources / Agentic AI / Core Features

Core Features

Agentic Studio

In-app configuration for the orchestration framework:

  • Models — OpenAI, Anthropic, or custom base URLs
  • Skills — reusable prompt + tool policy units
  • Agents — roles that bind model + skills + tools
  • Workflow packs — PDLC / SDLC / custom graphs with SVG editor
  • Clone / export / import — system packs are read-only; clone to customize

Graph workflows

Node types:

  • start / end — entry and exit
  • agent — runs ReAct with a bound agent (stage label optional)
  • gate — require artifact and/or explicit Advance
  • branch — condition on artifact presence or advance path

Seeded PDLC pack

Six stages with dedicated agents and artifacts (Discover → Iterate). Cyber Security and Product Research ship as thin stub packs.

ReAct agent core

  • Per-agent model binding via ModelRouter
  • Tools: list_files, read_file, write_file, edit_file, delete_file, run_project, run_code, search_files
  • Allowlists intersected across agent, skill, and node config

Workflow state

Per project:

  • pack_id, domain/slug, current_node_key, completed stages, node status

SSE streaming

Events: status, pack, node, agent, workflow, awaiting_gate, stage, content, tool_call, tool_result, files_changed, error, done.

Product page demo

Public demo at /products/agentic-ai against the seeded PDLC pack.