Agent-Native Toolkit

Use this skill when deciding whether app chrome, settings, collaboration, sharing, navigation, organization, setup, history, comments, or agent UX should be built app-locally or moved into reusable framework/toolkit pieces.

Core Rule

Apps own domain models, domain actions, and product-specific workflows. The framework and @agent-native/toolkit own repeated workspace behavior users expect to work the same everywhere.

Move behavior into shared toolkit primitives when it is:

• workspace-wide, such as settings, nav, search, org membership, or setup
• agent-visible, such as context, actions, run progress, or proof-of-done
• governed, such as secrets, permissions, sharing, audit, or billing
• repeated by two or more apps
• not tied to one domain model

Keep behavior app-local when the abstraction would hide important domain language or make a simple app-specific workflow harder to understand.

Agent Surface Contract

The repeated app shell has two distinct navigation surfaces:

• The left rail owns domain destinations and chat history when the app has a full-page chat route. Do not label a domain workflow as Chat just because the app was scaffolded from the chat template.
• Settings is not a rail item. It opens from the account menu (OrgSwitcher in the sidebar footer) and from ⌘, (Ctrl+, elsewhere), which AppProviders already binds. Keep the /settings route; don't add a Settings entry to items, secondaryItems, or the footer.
• The right AgentSidebar owns contextual agent work. Domain buttons that call sendToAgentChat should open it (openSidebar: true) so the user can see, steer, and review the agent without losing the page they were using.
• Keep / or /chat/* as the full-page chat surface when the starter provides one. Put domain workflows on named routes and wire the shell's route checks, navigation labels, and handoffs to those routes together.
• Use familiar message or neutral action icons for agent affordances. Never use sparkle, wand, magic, or robot icons as an AI label; the copy should carry the meaning.
• Give the right rail a quiet visual boundary with a subtle surface shift, divider, or both. The domain page and AgentSidebar should not collapse into a single undifferentiated background.
• Any button labeled as agent work must use sendToAgentChat with bounded context and openSidebar: true; local deterministic analysis should be labeled as local, preview, or analyze. For original/generated review, stack the source above the result by default and use side-by-side only for short, highly scannable content.
• Deterministic implementation does not make an AI-shaped experience deterministic. If the user expects research, analysis, generation, recommendation, synthesis, visible progress, or steering, route the button to the AgentSidebar and let the agent call focused actions. Keep revisions in the same thread instead of adding a second freeform prompt box.
• Standalone apps with AgentSidebar must use one AgentKit controller and transport. Core's AssistantChat export remains supported as an alias for AgentKitAssistantChat; use AgentKit slots and registries for custom message UI. Keep assistant-ui transcript/runtime imports inside the shared composer integration; when linked dependencies need Vite aliases, resolve one @agent-native/agentkit context and verify a real handoff.

Contextual agent UI is not a reason to expose every option at once. Start with the domain task's primary action, reveal review or configuration only when the current state needs it, and let the sidebar carry conversational depth.

AgentKit context launchers use a shared Core capability contract. Design and Slides consume get-agentkit-capabilities through useAgentKitCapabilities: Core filters provider APIs by the current app's exposed tool catalog and the same scoped connection grants or credential resolver used by provider requests. Figma context readiness is checked against its real Design processor. The shared hook also includes MCP servers only when they are connected and expose tools visible to the current request. Keep source retrieval (add-context) distinct from integration invocation intent (invoke-integration), and revalidate an invocation when it is read for submission. Keep only provider-specific pickers in app adapters; never duplicate readiness checks, infer agent-tool readiness from messaging connector status, or expose credential values. Keep a route to Integrations settings available when a connection is missing.

Visual Direction And Workspace Variety

Shared workspace behavior should be consistent without forcing every app into the same visual skin. Keep shell and component tokens semantic, then let each app declare a named direction in DESIGN.md before styling. A new app should choose its palette family and composition from the product context, compare nearby apps, and avoid inheriting their accent by default. Use the frontend-design visual-direction reference for mode, palette, type, density, shape, anti-references, and the distill / typeset / colorize / layout / polish / audit review vocabulary.

Do not make warm beige plus terracotta the workspace fallback. Preserve a workspace-level brand when one exists; otherwise keep shared chrome neutral and allow app-owned accents to distinguish products while retaining accessible semantic states and the shared AgentSidebar contract.

Discover Before Building

Before creating an app-local version of repeated workspace or agent UI:

1. Check the reusable kits below and the installed package documentation.
2. Search installed public components and source with docs-search and source-search.
3. Run agent-native eject --list to see the version-matched units published by the packages installed in this app.
4. Read customizing-agent-native and configure, compose, or eject the smallest unit instead of recreating shared behavior from memory.

Use public package exports at runtime. Published source and ejection manifests are discovery and ownership-transfer mechanisms, not private runtime APIs.

Design-System Boundary

Every app keeps an explicit design-system seam in app/design-system.ts using defineDesignSystem from @agent-native/toolkit/design-system, and supplies it to ToolkitProvider. The semantic contract contains:

• nine leaf components: ActionButton, IconButton, TextField, TextArea, Spinner, Skeleton, Status, Surface, and Avatar
• eight behavior components: Tooltip, Menu, Popover, Dialog, Picker, Checkbox, Switch, and Tabs

These are semantic contracts, not styling contracts. An adapter may use Tailwind/shadcn, MUI-style theme providers, React Aria, CSS modules, CSS-in-JS, or another React design system. Do not assume CVA, utility classes, or even a className; behavior adapters may supply their overlay and focus implementation wholesale while honoring portal, focus-restoration, keyboard, dismissal, ARIA, and z-index interoperability.

Pages, routes, and domain components import ordinary controls through the app's local adapter layer, usually @/components/ui/*. They must not import @agent-native/toolkit/ui/* directly. Toolkit feature exports are still the right home for shared workspace behavior; their presentation flows through the registered semantic components, feature controller, and product-level slots.

Customer adapter packages are normal npm packages imported explicitly by the app. Never auto-detect them or load React components from JSON. Run the adapter against @agent-native/toolkit/conformance in customer CI before adopting it.

Settings Direction

Durable settings belong in Settings. The agent sidebar should not become a second settings app; it can show contextual quick controls and deep links. Settings has the same groups in every app, and page ids are stable URL segments (/settings/<page>/<sub>):

• Account: profile, preferences, security
• Connections: integrations (integrations/builder), api-keys
• Agent: model, instructions, memory, skills, files, sub-agents
• Organization: org, members, usage, and for owners and admins auth, apps, infra, audit
• The app's group: app (areas at app/<id>), notifications, automations, channels (channels/<platform>), mcp, creative-context
• Footer: labs, whats-new

Link with buildSettingsRoute(page, sub?, { anchor? }); old tab and section ids resolve through the redirect table in packages/core/src/navigation/settings-redirects.ts, the only place that maps them. Account › Profile is the canonical profile surface (get-user-profile, update-user-profile); don't build an app-local profile page.

When adding a new API key, OAuth grant, provider connection, model selector, app preference, notification preference, or usage/billing surface, find the page that owns that kind of setting first: provider keys go through the one provider dialog on Model, other keys on API keys, channels on Channels, and app-only preferences in the app's group. Only add sidebar UI when it is needed in the moment of agent use.

The app's group in Settings

Settings has a group named after the app. Core owns its pages: General, Notifications, Automations, Channels, MCP server, Creative context, plus Labs and What's new in the footer. A template supplies only its own content, through these SettingsTabsPage props:

• generalGroups: the app's own SettingsGroups on its General page. Core puts Agent › Default model above them (owners and admins change it; the agent uses manage-agent-engine set-app-default) and This browser › Demo mode below. Until a template passes it, today's general shows there.
• appAreas: [{ id, label, content, visible?, keywords?, searchEntries? }], tabs on the General page routed /settings/app/<id>. Set visible: false while the lab behind an area is off. A tab with settingsPlacement: "app-area" in extraTabs works the same.
• notifications (plus notificationsSearchEntries): the Notifications page shows only when this is passed.
• mcpAbout: the MCP server page's about line, naming what an MCP host can do in this app.
• Labs come from labs; What's new comes from whatsNewMarkdown, or from the ChangelogSettingsCard passed as whatsNew.
• Any other extraTabs item becomes its own page in the app's group. To swap a core page for a variant, registerSettingsPages([{ ...corePage, component }]) at module scope (Dispatch's Members keeps its app-role column this way). To add a group to one channel's page, registerChannelSettingsExtensions.

The route needs a settings.$.tsx splat next to settings.tsx, and the layout hides its own sidebar and header on Settings routes while the flag is on or loading (isSettingsPathname), because the shell brings its own.

Search entries per area: each searchEntries item's hash is the row's SettingsRow id, and a hit opens that area's tab and scrolls to the row. With the flag off, the same props render as today's tabs, so a migrated template works either way. Link with buildSettingsRoute("app", "<area>"), never a hand-written path.

Gate UI on a lab with useLab(LAB_DEFINITION) rather than useLab(key): a definition reads as its defaultEnabled until the server answers, while a bare key reads as on.

Integration Setup Preflight

Before building any setup, settings, credential, OAuth, or connection surface, search the workspace/provider connection catalog first. If the provider already has a reusable connection, use its catalog, app grant, and scoped credential resolver rather than registering a parallel secret. Only then classify fields that still need app-local setup by lifecycle and scope:

Do not register every provider field as a generic secret, mark every field as required, or create a second credential-management surface. One logical connection should normally produce one onboarding outcome. A custom setup page is appropriate only when it adds domain-specific guidance or readiness checks; it should link to or call the shared settings, OAuth, and action surfaces rather than duplicating their storage or transport.

Reusable Kits

• Settings kit: one searchable Settings page with Account, Connections, Agent, and Organization groups plus a group for the app's own settings. The index is built from page declarations; give every row a search entry whose anchor is its SettingsRow id so users find settings by name.
• Collaboration kit: Yjs docs, presence, agent presence, live cursors, remote selections, recent edit highlights, real-time sync indicators, and undo/redo grouping.
• Sharing kit: private/workspace/org/public-link access, invites, roles, expirations, agent-readable links, and resource registration.
• Navigation and command kit: app shell, side nav, breadcrumbs, app switcher, command palette entries, recent resources, pinned resources, and global search.
• Organization kit: folders, tags, favorites, archive, trash, ownership, membership, and common resource metadata.
• Setup and connections kit: declarative setup requirements, model readiness, missing-secret states, OAuth grants, and provider connection health.
• Agent UX kit: sidebar, composer, staged context, mentions, voice, human approval, generative UI, progress, and screen-state exposure.
• Custom block kit: optional extension creation, viewer chrome, slots, and promotion affordances. Apps must opt in; Core keeps the sandbox, SQL storage, access checks, and compatibility routes, while Toolkit owns reusable adoption UI. Analytics may expose this as one-off Custom Blocks; durable behavior should be promoted to app code.
• Chat history kit: presentational chat lists and recent-chat rails belong in Toolkit; Core keeps thread persistence, agent execution, transport, and page-to-sidebar handoff. Use Toolkit's ChatHistoryRail for the standard five-item sidebar preview and a footer row with New chat followed by an ellipsis disclosure up to fifteen. Apps inject routing, labels, and domain actions.
• Data grid kit: provider-agnostic spreadsheet mechanics belong in @agent-native/toolkit/data-grid. Apps provide rows, typed columns, editor slots, selection and width state, persistence callbacks, and product-level row/body slots. Keep database models, access checks, grouping, drag/drop, and domain actions in the app adapter.
• Agent page kit: the full-page /agent surface (AgentTabsPage from @agent-native/core/client) with Context, Files, Connections, Jobs, and Access tabs plus a Personal/Organization scope toggle. The canonical home for context transparency, MCP servers, A2A remote agents, recurring jobs/automations, and external-client connect flows. See the agent-page skill.
• History and recovery kit: audit log, activity feed, version history, checkpoints, undo, redo, restore, and proof-of-done.
• Comments and review kit: anchored comments, pins, mentions, review requests, resolved threads, agent follow-up tasks, and notifications.
• Workflow and observability kit: notifications, approvals, scheduled work, background runs, recurring jobs, traces, evals, feedback, and run timelines.

Implementation Checklist

When adding or refactoring one of these areas:

1. Search existing framework and template code for duplicated UI or actions.
2. Decide the shared contract: data shape, action API, feature-level headless controller, default view, semantic components, and product-level render slots.
3. Keep shared data provider-agnostic and scoped by auth/sharing rules.
4. Expose the same capability to the UI and agent through actions or documented client helpers.
5. Register app-specific labels, routes, resource adapters, and settings panels instead of hardcoding app names in core UI.
6. Update docs and relevant skills so future apps discover the shared path.
7. Keep the component easy to adopt piecemeal: expose props/slots first and ship readable source plus a complete ejection unit so apps can take ownership of the smallest feature when needed. See customizing-agent-native for the configure → compose → eject → propose seam ladder.
8. Keep one controller as the source of truth for the default and custom render paths. A custom design must not fork actions, analytics, async state, or accessibility behavior.
9. Verify the default adapter and at least one non-Tailwind adapter with the conformance kit, including focus and portal stacking across mixed overlay implementations.

Related Skills

Read these alongside this skill when the work touches the specific area:

• sharing
• real-time-collab
• real-time-sync
• client-side-routing
• context-awareness
• onboarding
• secrets
• audit-log
• observability
• frontend-design