Tool reference
Typed read and conversational tools exposed by the Vidext Operator MCP server.
The endpoint exposes these public tools. None of its public tool names contain _vidext_.
| Tool | Access | Purpose |
|---|---|---|
search_knowledge | Read | Search connected knowledge available to the caller. |
read_knowledge_source | Read | Read one accessible source returned by search. |
list_modules | Read | List modules visible to the caller. |
get_module_status | Read | Inspect a module's authoritative outline and build status. |
get_operation | Read | Poll an asynchronous operation by moduleId and operationId. |
consult_operator | Read | Ask the model to reason over a server-enforced read-only tool surface. |
list_operator_chats | Read | List up to 50 recent chats for the authenticated user and organization. |
read_operator_chat_history | Read | Read the bounded history of an accessible participant chat. |
ask_operator | Write | Start or continue a write-capable operator conversation. |
Conversational tools
consult_operator accepts a prompt but cannot mutate state: the server restricts the model to read-only tools. ask_operator can plan, build, and edit. It requires an idempotencyKey on every call so retries do not duplicate write-capable work. It can also accept a chatId, activeModuleId, and locale. Do not treat internal Operator tool names as callable MCP tools.
When ask_operator requests user input, relay the complete question and options, then continue the same chatId. Its response includes correlation IDs and, for asynchronous work, a typed next action:
{
"tool": "get_operation",
"arguments": { "moduleId": "018f1111-2222-7333-8444-555555555555", "operationId": "operation_..." },
"pollAfterMs": 1500
}Wait at least pollAfterMs, call get_operation with the supplied arguments, and repeat while its result has terminal: false. A nonterminal result supplies the next delay; a terminal planner result can supply the builder as nextAction, which you should follow.
Last updated on