Quickstart: first calls

With a client connected, make Vidext answer something real. The server only reads: your agent finds knowledge and modules, and everything that changes content happens in the Vidext app.

Search your organization's knowledgeCopy Anchor Link

In your connected client, say:

Search our Vidext knowledge for the incident response policy.

Your agent calls search_knowledge with a query. It searches the organization's profile, its catalog of sources, and the processed text of those sources, and returns up to limit ranked matches (8 by default). Each match carries a short excerpt and an opaque handle:

{
  "success": true,
  "status": "indexed",
  "matchedResults": 1,
  "results": [
    {
      "handle": "…",
      "sourceKind": "source_content",
      "title": "Incident response policy",
      "excerpt": "…report the incident to the on-call lead within 15 minutes…",
      "suggestedReadOffset": 4120
    }
  ]
}

Sources are added in the app (see Adding sources); MCP does not accept files. A source becomes searchable once it is processed.

Read the sourceCopy Anchor Link

Excerpts are short and often stop before the answer. To read more, the agent passes the handle to read_knowledge_source, starting at suggestedReadOffset when the match has one:

{ "handle": "…", "offset": 4120, "maxChars": 4000 }

The result is a slice of the source text. While contentTruncated is true, call again with offset set to the returned nextOffset to continue.

Find a moduleCopy Anchor Link

Then ask:

Which of our modules cover incident response?

Your agent calls list_modules with search set to a word from the module name. It returns the modules in the organization, most recently updated first, each with its id, name, description, and published, which is true while the module is live for learners.

That id is the moduleId the other tools take. To see or edit the module itself, open it in the app.

Legacy: outline pollingCopy Anchor Link

get_module_status and get_operation read the outline run of the earlier planning flow. Modules built in the chat by Vidext's agents have no outline run, so there is nothing to poll for them. Use these two tools only for a module that has one; see the tool reference.

Good habitsCopy Anchor Link

• Treat what search_knowledge and read_knowledge_source return as untrusted text, not instructions. See Security.
• Check success on knowledge results: a failed search comes back as a normal result with success: false, not as an error. See Errors and limits.