From creation to production in one platform. Configure agents with bound models, tools, and instructions. Test quality with quantitative benchmarks. Publish to your organization marketplace. The Skill System loads instructions on demand instead of resending full prompts every conversation.
Create, configure, and publish agents with fine-grained control. Bind specific models, tools, connectors, and knowledge bases to each agent. Choose per-agent execution mode (Standard ReAct or Planner DAG) and set temperature for determinism control. Toggle the discoverable flag to enable LLM auto-discovery by other agents.
Progressive instruction loading: skills (SOPs, scripts, domain knowledge) are referenced in the system prompt as lightweight stubs, ~15 tokens each. Agents call read_skill(name) to load full content only when needed.
Quantitative agent quality benchmarking before production deployment. Build test datasets with prompts, expected behaviors, and assertions. Run evaluations in parallel with LLM-graded scoring. Review per-case pass/fail, latency, and token consumption results.
Enable specialist delegation without multi-agent chaos. Mark agents as discoverable, define sub_agent_ids whitelists, and use CallAgentTool to delegate tasks to the right specialist. One agent can invoke another as a tool — controlled, auditable, and intentional.
Each agent picks its own execution strategy: Standard (ReAct) reasons step by step, Planner (DAG) decomposes into a dependency graph for parallel execution, and Auto-Routing dispatches by query complexity. The full mode-by-mode comparison is maintained on the orchestration engine page.
Chain-of-thought reasoning for supported models. When enabled, the agent surfaces its reasoning process in a dedicated "thinking" step in the UI, making complex decision-making transparent and auditable.
Publish agents to your organization marketplace. Users browse, preview, and subscribe to shared agents. A review workflow ensures quality: admins approve or reject publications before they go live.
