Skip to main content
The Provider Directory lists the systems Parable can connect to and the Provider Plugins available for each system. Browse it before creating a connection so your integration can select a supported grouping and authentication strategy. In the Workspace Providers page, search matches Provider names across Needs Attention, Required, Suggested, Connected, and All providers. Counts reflect the matching Providers; empty sections disappear with their headings and spacing. Search also filters requested Providers by name. Clearing search restores the directory. Catalog search resets to the first page and fetches results without reloading connection state or the page shell. Status sections and All providers use the same grid spacing. Provider card headers reserve room for a title and two caption lines, including when the caption is empty or only one line. Longer descriptions and status reasons are clamped to two lines, keeping the cards the same height. Captions use PP Fraktion Sans at 12px. Status pills stay beside the Provider name, centered against its capital-letter top and baseline, with or without a caption. The title row and caption are centered together against the logo and card. Descenders extend below the baseline. Font space is trimmed using font metrics, without a fixed positional offset. Pills retain their full label while long names give up horizontal space first. Long Provider names truncate horizontally while accents and descenders remain visible. An idle admin card does not repeat “Connected” beneath its badge; useful scheduling copy still appears when available. Customer idle cards keep their caption blank.

​
Browse Providers

The Workspace API has no public address yet. Browse Providers and your connections from Providers in the Workspace app. See Availability.
Directory responses distinguish several layers: Leaf groupings expose the Provider Plugin choices that can be configured. A Provider may have more than one grouping when its products or deployment modes require different credentials or endpoints. Open Providers directly from the Workspace sidebar, which also switches to People and Protection without a duplicate tab bar above Providers. Use the directory filters to browse available Providers, connected services, connections needing attention, and requests. Provider details and connection forms retain their own headers and breadcrumbs. The connection documentation rail contains Overview, Setup Guide, and Troubleshooting alongside the configuration form. Open Artifacts from the directory to inspect uploaded data assets.

​
Inspect existing connections

List the current Workspace’s Provider connections:
Each instance includes its connection state, its grouping and Provider, and its Provider Plugin with the available authentication strategies. Connection lists leave out Provider docs and data streams: no grouping or authentication strategy in an item carries docs, and connector.taps is empty unless the request adds ?includeTaps=true, which returns every data stream with its schema. Read one connection by ID (/api/vendors/tenant-connector-instances/{id}) for its docs and data streams. The same rule applies to the per-Provider, per-grouping, and grouped lists. Next, create a Provider connection.

​
Frontend cache isolation

Each server-rendered request owns its QueryClient. Browser navigation retains one client within the resolved Workspace, authenticated session, effective user and permissions. A scope change replaces the client and its observers before rendering the new scope. Session tokens are represented by one-way fingerprints, never serialized as credentials. Server clients do not schedule browser GC timers. The QueryClient instance is the authenticated cache namespace, so existing query keys, reads, writes and invalidation prefixes stay unchanged. All consumers use that provider-owned client. The audit covers the Provider catalog and exclusion filter, connections, requests, vendor/grouping details, validation polling, admin Parables and their refs/packages/deployments/report context, user search and Workspace search/detail. Search, pagination and resource identifiers remain in their existing keys; polling and cache defaults are unchanged within a scope. Persisted command receipts include scope in their own keys. The former sidebar hover-prefetch Map has no remaining consumers on current main. Existing user/resource-scoped draft recovery and actor-scoped preferences are unchanged. No TanStack persistence or dehydration path exists in the app; loaders supply initial data. Regression coverage includes sequential and overlapping streaming SSR, distinct authenticated scopes, stable browser caching, scope transitions, late responses, hydration, scoped invalidation, polling and receipt reattachment. Local synthetic verification compares Tech-startup and Acme’s initial HTML with JavaScript disabled, then checks hydrated membership and supported login, Workspace and impersonation transitions. Browser request concurrency is complemented by a deterministic streaming barrier in the regression tests.