Installed extensions

extensions.catalog.read lists what your organization has installed: each extension, its products, the tools those products offer and the models it serves.

import { createHttpPlatformTransport, createScrydonSDK } from '@scrydon/sdk/platform'

const platform = createScrydonSDK(
  createHttpPlatformTransport({
    baseUrl: 'https://scrydon.com',
    headers: () => ({ authorization: `Bearer ${accessToken}` }),
  })
)

const catalog = await platform.extensions.catalog.read({ capability: 'llm' })
for (const product of catalog.products) console.log(product.name, product.toolIds)

Over HTTP it needs tools:read and an organization-level caller. It is read-only and safe to retry. enabledOnly defaults to true, which lists only what the organization has enabled; enabledOnly: false lists everything installed. extensionId and capability narrow the listing.

From a Scrydon tool: ctx.extensions

Inside a Scrydon tool the same operation is ctx.extensions.catalog.read, and it answers for the run's own workspace: it lists only what is enabled for the organization and available to that workspace. An extension the organization enabled for another workspace only is not listed.

const catalog = await ctx.extensions.catalog.read({ capability: 'llm' })

On this path enabledOnly is always true. A request with enabledOnly: false, or one that names a workspace, is refused before anything is read, so a tool cannot list another workspace's extensions or everything installed in the organization. Declare "extensions.catalog.read" in the tool's platformOperations.

In a workflow, select the Scrydon tool List installed extensions in an Agent block's tools. It returns names and ids — extensions, products, tools and models — and never a tool's input or output schema. Cortex runs it without asking for approval, because it only reads, and nothing is sent outside the platform.