> ## Documentation Index > Fetch the complete documentation index at: https://docs.orbit.devotel.io/llms.txt > Use this file to discover all available pages before exploring further. # The Agents workspace map > One orientation page for the /agents hub: the card row, the tabs, the cross-hub routes, the old-URL redirects, and which workspace role sees which surface. # The Agents workspace map `/agents` is the home of your AI workspace in the Orbit dashboard. Everything you do with AI agents (building, grounding, monitoring, reviewing transcripts) either lives on this page or hangs off it as a dedicated route. This guide is the orientation: what each card and tab is for, when to open it, where the older URLs now go, and which workspace role sees what. For agent concepts and the create-to-deploy lifecycle, use [AI agents on Orbit](/agents/overview). For the rest of the dashboard, see the [dashboard map](/concepts/dashboard-map). ## The card row Across the top of `/agents`, a row of cards links the sibling surfaces that are full workflows of their own rather than tabs inside the hub. The current cards, left to right: | Card | Routes to | Open it when | | - | - | - | | Voice Agent Templates | `/agents/templates` | You want a working voice agent in one click (cart recovery, appointment reminders, lead qualification) instead of configuring from scratch. | | Model Presets | `/agents/model-presets` | You want a one-click transcriber + LLM + voice bundle (Balanced, Fast, Cost Saver) and want to compare latency, cost, and quality before committing. | | Prompt Template Library | `/agents/prompt-templates` | You want to fork a curated, versioned system-prompt starter into a new agent. | | Automation Opportunity Miner | `/agents/automation-opportunities` | You want your transcripts mined for what to automate next, ranked by projected deflection and ROI. | | Guardrail Effectiveness | `/agents/guardrail-analytics` | You want firing rates, violation trends, and per-policy effectiveness on live traffic, to check that your guardrails actually catch violations. | | RAG Grounding Citations | `/agents/grounding-citations` | You want to audit whether your voice agents' answers carry knowledge-base citations on live calls. | | Squads | `/agents/squads` | You group agents into teams that hand off to each other automatically. | | Knowledge Base | `/agents/knowledge-base` | You ingest and manage the RAG library that grounds your agents' answers. | | Contact Memory Store | `/agents/memory` | You browse or edit long-term memory per contact, across all agents. Per-agent memory lives on each agent's own Memory tab; this card is the tenant-wide store. | | Flows | `/flows` | You build no-code, cross-channel automation that agents can participate in. | | Custom Tools | `/agents/custom-tools` | You register HTTPS webhook tools, with JSON-Schema input contracts, that any agent can call. | The cards are discoverability hops: each target has its own full console, listed again in the [cross-hub routes](#cross-hub-routes) table below. ## The four canonical tabs Below the card row, the page is a tab strip. Four tabs do the day-to-day work: | Tab | What it shows | When to open it | | - | - | - | | **Agents** | Your agent grid: name, type, status, deployed channel, conversation counts, and token spend against caps, with an at-a-glance live strip above it. | The default view. Create, test, deploy, and manage agents here. [Creating agents](/agents/creating-agents) | | **Studio** | The visual canvas for composing and iterating on an agent. | When you build or debug an agent's behavior graphically instead of field by field. [Agent Studio canvas](/guides/agent-studio-canvas) | | **Marketplace** | Pre-built agents you can install into your workspace. | When you want a starting point maintained by Devotel rather than a blank agent. [Marketplace](/agents/marketplace) | | **Conversations** | The historical transcript browser for agent conversations. | When you review what your agents actually said, how calls resolved, and where they struggled. | A fifth tab, **Live**, is the deep drill-down for in-flight conversations. It is shown only to `owner`, `admin`, and `supervisor` roles; other roles never see the tab trigger, and navigating to a live conversation directly is refused for them. See [Roles on this workspace](#roles-on-this-workspace). Tabs are addressable: selecting one sets `?tab=` in the URL, so `/agents?tab=conversations` opens the hub with Conversations preselected. Use these URLs to point a teammate at an exact surface. ## Cross-hub routes Every card and tab has a canonical route, and the hub's deeper surfaces each have one too. Route for route: | Route | Surface | Go deeper | | - | - | - | | `/agents/studio` | The canvas builder (Studio tab). | [Agent Studio canvas](/guides/agent-studio-canvas) | | `/agents/marketplace` | Installable pre-built agents. | [Marketplace](/agents/marketplace) | | `/agents/squads` | Multi-agent teams with automatic handoff. | [Squads](/agents/squads), [squad guide](/guides/agent-squads) | | `/agents/knowledge-base` | The RAG library that grounds answers. | [KB document inspector](/agents/kb-document-inspector) | | `/agents/memory` | Tenant-wide contact memory browser. | [Contact memory](/guides/agents-memory), [memory administration](/agents/memory-administration) | | `/agents/custom-tools` | Webhook tool registry any agent can call. | [Custom tools console](/guides/agents-custom-tools) | | `/agents/templates` | One-click pre-built voice agents. | [Intro to voice AI agents](/agents/intro-to-voice-ai-agents) | | `/agents/model-presets` | One-click transcriber + LLM + voice bundles. | [Model presets](/guides/agents-model-presets) | | `/agents/prompt-templates` | Shared, curated, versioned prompt library. | [Prompt template library](/guides/agents-prompt-template-library) | | `/agents/ambient-outreach` | Proactive outreach runs handled by agents. | [Ambient outreach](/agents/ambient-outreach), [guide](/guides/agent-ambient-outreach) | | `/agents/automation-opportunities` | Transcript-mined automation candidates. | [Automation opportunities](/agents/automation-opportunities) | | `/agents/from-prompt` | Generate a first-draft agent from a plain-language prompt. | [Build an agent from a prompt](/guides/agent-from-prompt) | | `/agents/grounding-citations` | Citation audit for voice answers. | [Grounding citations](/guides/agent-grounding-citations), [concept](/agents/grounding-citations) | | `/agents/guardrail-analytics` | Guardrail firing and violation analytics. | [Guardrail effectiveness](/agents/guardrail-effectiveness) | | `/agents/tool-approvals` | Approve or reject agent tool calls awaiting review. | [Tool approvals](/guides/agent-tool-approvals) | | `/agents/live` | In-flight conversation drill-down (role-gated). | [Live monitor](/agents/live-monitor) | ## Old URLs and redirects Three legacy addresses forward to the hub automatically, so old bookmarks and deep links keep working: | You visit | You land on | | - | - | | `/ai` | `/agents` | | `/ai/memory` | `/agents/memory` | | `/agents/conversations` | `/agents?tab=conversations` (the Conversations tab inside the hub) | There is no standalone `/ai` or `/agents/conversations` page anymore; the hub and its tabs are the canonical surfaces. Bookmark the targets. ## Roles on this workspace | Role | On `/agents` | | - | - | | `viewer` | Read-only: the grid, tabs, and transcripts are visible; create, edit, and deploy controls are not. | | `developer` | The build surfaces: Studio, prompt and model preset editing, custom tools, and deployments. | | `supervisor` | The monitoring seat: the Live tab and the live conversation drill-down. | | `owner`, `admin` | Everything on the workspace, including Live and forking from the shared template libraries. | The console hides tabs and controls your role cannot use, so a missing tab is usually a role question, not a bug. The hiding is a courtesy: the API enforces the same rules server-side, so a direct link to a gated surface returns 403 regardless of what the navigation shows. ## Troubleshooting * **A tab or control I expect is missing.** Check your workspace role against the table above. The Live tab in particular only renders for `owner`, `admin`, and `supervisor`. Ask an owner or admin to raise your role if the job needs it. * **I followed a deep link and got a 403.** The navigation hide is UX; the surface enforces roles server-side. This happens when a link to a gated route such as `/agents/live/...` is shared with a role that cannot open it. * **An old bookmark lands somewhere unexpected.** The legacy addresses in [Old URLs and redirects](#old-urls-and-redirects) forward automatically; if a link predates them, use the `?tab=` form, for example `/agents?tab=conversations`. * **A tab's count badge says "Couldn't load this count".** The badge's count fetch failed transiently; the tab itself still works. Refresh, and if the badge keeps failing while the tab loads, your session's read scope for that count is the thing to check. * **Agent runtime errors (422s, 429s, quota codes).** Those are API-side, not navigation-side; work through the [agent error reference](/troubleshooting/agent-errors). ## Next steps * [Creating agents](/agents/creating-agents): the full create-to-deploy walkthrough. * [Agent Studio canvas](/guides/agent-studio-canvas): build on the Studio tab. * [Squads](/guides/agent-squads): multi-agent handoff teams. * [Contact memory](/guides/agents-memory): the memory surfaces behind the Contact Memory Store card. * [Marketplace](/agents/marketplace): install pre-built agents. This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.