# AI Agents components

> 9 AI Agents components for React, shadcn, Vue 3, Laravel Blade and HTML + Alpine.js: AgentPersonaEditor, Agent Steps and Confirm, AiModelPicker, AiUsageCost, BrainList, KnowledgeGaps, ModelRoutingEditor, ResearchRun, Section Board.

Source: https://docs.nasaqui.com/components/groups/ai-agents

- [AgentPersonaEditor](https://docs.nasaqui.com/components/agent-persona-editor): Editor for an AI agent persona with name, tagline, colour, icon, traits, model, greeting and a markdown persona with write and preview tabs, section chips, counts, save and revert.
- [Agent Steps and Confirm](https://docs.nasaqui.com/components/agent-steps): The tool calls an agent made and a human-in-the-loop confirm before it acts. A step list with arguments, results and retry, plus a diff review with per-change ticks, risk, and Apply or Reject.
- [AiModelPicker](https://docs.nasaqui.com/components/ai-model-picker): Choose the model, its reasoning effort and a required agent or skill, as cards or one compact toolbar row, plus a PersonaPicker with prompt starters and the compact AiModelSelect that CopilotChat uses.
- [AiUsageCost](https://docs.nasaqui.com/components/ai-usage-cost): AI spend overview with total, token, billed, unbilled and marked-up client price tiles, a daily stacked cost chart and a by-model, product or run breakdown with token columns, plus a TokenCostMeter for one run.
- [BrainList](https://docs.nasaqui.com/components/brain-list): Zekra-style list of AI brains as a table or cards with status, access, memory and source counts, members, tags and last activity, plus search, filters, bulk select and loading and empty states.
- [KnowledgeGaps](https://docs.nasaqui.com/components/knowledge-gaps): Queries a knowledge base could not answer, grouped as open, indexed or dismissed, with hit counts, first and last asked, and actions to mark indexed, dismiss or reopen.
- [ModelRoutingEditor](https://docs.nasaqui.com/components/model-routing-editor): Admin editor for AI model routing with an automatic-routing switch, a task class to model and fallback route table, an active backend switch and provider or node registration with modality chips.
- [ResearchRun](https://docs.nasaqui.com/components/research-run): A research run with an ask box, live stage progress, an answer whose sentences cite numbered evidence, the evidence list, sources, confidence and failed or cancelled states with retry.
- [Section Board](https://docs.nasaqui.com/components/section-board): An ordered set of sections, each generated by a prompt, a model and a few settings. View mode shows the content; edit mode reorders by drag or keyboard and edits each section in a dialog.