Search the Workday developer documentation by concept, widget, API, or task.

Returns lightweight pointers (title, breadcrumb, doc_id, link) — NOT full page text. Call get_workday_doc with a returned doc_id to read a page.

When the full-text index is built (npm run build-index), this searches page BODIES (BM25), so intent/how-to phrasing works well — e.g. "sortable grid with pagination", "format a date in a PMD expression". Without the index it falls back to TITLE-only matching (engine="title"), where specific how-to phrasing may return weak hits — build the index for best results. The response 'engine' field tells you which mode ran.

Args:

• query (string): concept/widget/API/task, e.g. "validate a field on submit"

• section ('Extend Apps' | 'Integration Apps' | 'Workday APIs' | 'Workday Developer Copilot'): optional filter

• limit (number): max results, 1-50 (default 10)

Returns: { query_echo, engine, count, results: [{ doc_id, title, breadcrumb, html_url, score }] } Higher score = stronger match. Empty results means no match — try broader terms or browse_workday_toc.

Fetch the full markdown for one Workday documentation page.

Accepts EITHER a doc_id (from search_workday_docs) OR a developer.workday.com "/wcp_docs/.html" link. Always returns the canonical html_url so you can cite/link the page even if the body can't be fetched.

Args:

• doc_id (string): a doc ID from search_workday_docs

• html_url (string): alternatively, a developer.workday.com/wcp_docs/.html link

Returns: { doc_id, html_url, truncated, markdown, error? }

• markdown: the page body (string), or null if the body fetch failed (see error)

• truncated: true if the body exceeded 60000 chars and was cut

• error: present only when the body could not be fetched (html_url is still valid to link)

Explore the documentation hierarchy when keyword search misses.

Pass a breadcrumb path to expand a node; an empty path lists the top-level sections (Extend Apps, Integration Apps, Workday APIs, Workday Developer Copilot, ...). Use the returned doc_id values (non-null nodes are real pages) with get_workday_doc.

Args:

• path (string[]): breadcrumb to expand, e.g. ['Extend Apps','Extend App Components Reference'] (default [])

• depth (number): how many levels deep to expand, 1-3 (default 1)

Returns: { path, nodes: [{ title, doc_id, child_count, children? }] }

Ground Extend app-building in the authoritative reference. Given what you're trying to build or use, this returns the BEST-matching reference page WITH its content inline (excerpt) — not just a pointer — plus alternative pointers. Use this before writing PMD/widget/scripting code so you ground on the real schema/signature instead of guessing.

Scoped to the "Extend App Components Reference" (PMD tags/widgets, model components, PMD functions, PMD scripting). For broader/conceptual docs use search_workday_docs. Best results require the full-text index (npm run build-index).

Args:

• intent (string): what you want to build/use, e.g. "sortable grid with row actions", "function to format a date", "validate a field before submit"

• kind ('any' | 'component' | 'function' | 'scripting'): narrow the scope (default 'any') component = PMD tags/widgets + model components; function = PMD functions; scripting = PMD scripting tags

• alternatives (number): how many additional pointers to return, 0-10 (default 4)

Returns: { intent, kind, engine, best: { doc_id, title, breadcrumb, html_url, excerpt, truncated } | null, alternatives: [{ doc_id, title, breadcrumb, html_url, score }] } If best is null, nothing matched the scope — widen kind, rephrase, or use search_workday_docs.