Documentation Tools

MCP tools for searching and reading B2C Commerce documentation across multiple corpora: Script API reference (dw.* classes/modules), Developer Center guides (conceptual/how-to content), Salesforce Help (administration and merchandising content), tooling documentation (CLI/MCP/SDK guides), standard job steps, and XSD schemas. All corpora are indexed locally (the index ships with the SDK); Script API, Developer Center guides, Salesforce Help, and tooling docs fetch their full text online when read (cached locally after the first read), while job-step content is bundled. No instance configuration or authentication is required. Available in every toolset.

By default the tools return compact results — docs_search returns a short result list with summary fields, docs_list returns a table of contents, and docs_read returns a bounded slice of long documents. Each accepts options to return more (verbose, limit/offset, maxLength).

Restricting available topics: Launch the server with --docs-topics <categories> (or the SFCC_DOCS_TOPICS env var) to bound the entire documentation corpus to a comma-separated allowlist (e.g. --docs-topics sfnext,commerce-api,tooling or --docs-topics help-admin,script-api,tooling for an admin-focused server). This is a hard configuration boundary — the category parameter is narrowed to the allowlist, per-call category/workspace narrow within it, and docs_read will not resolve an id outside it. Unknown category names are ignored with a warning. The CLI commands (b2c docs search/read) support the same restriction via the --topics flag or SFCC_DOCS_TOPICS env var; all three surfaces also read the docsCategories config field (dw.json docs-categories, SFCC_DOCS_CATEGORIES, package.json). Use it to expose only the docs relevant to a given project.

---

Documentation (Multi-Corpus)

docs_search

Content-aware search across all corpora using BM25-style ranking. Results include the key fields by default (id, title, category, summary, score); pass verbose=true for keywords and url. The MCP server auto-detects the workspace's storefront framework at startup and uses it to weight results.

Parameter	Type	Required	Default	Description
query	string	Yes		Search query (keywords, class name, or topic)
limit	number	No	5	Maximum number of results
verbose	boolean	No	false	Include keywords, url, and sourceUrl on each result (larger payload)
category	string	No	(all)	Filter by category: script-api, commerce-api, pwa-kit-managed-runtime, sfnext, sfra, b2c-commerce, tooling, job-step, help-admin, help-merchant
workspace	string	No	auto	Workspace awareness: auto (auto-detected), all (no preference), or specify one or more types comma-separated: cartridges, sfra, pwa-kit-v3, storefront-next

Returns: {query, category?, workspace?, results: [{id, title, category, summary?, score, keywords?, url?, sourceUrl?}]}. Higher score = better match (results are pre-sorted best-first). keywords, url, and sourceUrl appear only when verbose=true. The url field is the .html page for opening in a browser, while sourceUrl is the raw .md source. Use summary to identify the right entry, then read it with docs_read.

Workspace awareness: By default, the tool uses the auto-detected workspace (shown in the tool description) to boost relevant documentation. A detected or specified workspace boosts relevant categories (x1.4) and de-boosts competing storefront framework guides (sfra/pwa-kit-managed-runtime/sfnext that are not the current workspace) by x0.3. It never hides/filters — only reweights. To hard-scope results to a category, use the category parameter.

Category boost mapping:

• Always relevant (any workspace): b2c-commerce, tooling
• cartridges → script-api, job-step
• sfra → sfra
• pwa-kit-v3 → pwa-kit-managed-runtime, commerce-api
• storefront-next → sfnext, commerce-api

SFRA vs cartridges distinction: A SFRA project is detected as both cartridges and sfra (SFRA is a refinement — it has the app_storefront_base cartridge). A cartridge project that is not SFRA (custom API cartridges, or a PWA Kit / Storefront Next repo that also ships cartridges) is cartridges only — so it will not boost sfra docs.

Examples:

• Search with auto-detected workspace: docs_search(query="components")
• Search all docs (no workspace preference): docs_search(query="passwordless login", workspace="all")
• Filter by category: docs_search(query="passwordless login", category="commerce-api")
• Find Script API classes: docs_search(query="ProductMgr", category="script-api")
• Search with specific workspace types: docs_search(query="hooks", workspace="sfra,pwa-kit-v3")

docs_read

Read documentation for a specific entry. Script API, Developer Center guides, Salesforce Help, and tooling docs fetch content online from the sourceUrl (.md) and cache it locally after the first read, with offline fallback (summary + headings + both URLs). Job-step content is bundled (no fetch). Long documents are truncated to maxLength characters — page through the rest with offset when truncated is true.

Parameter	Type	Required	Default	Description
query	string	Yes		Entry ID (e.g., "dw.catalog.ProductMgr", "sfnext/sfnext-get-started", "commerce-api/slas-passwordless-login-registration") or fuzzy query
offset	number	No	0	Character offset to start reading from (for paging long docs)
maxLength	number	No	12000	Maximum characters of content to return

Returns: {entry: {id, title, category, summary?, keywords?, url?, sourceUrl?, filePath?, preview?}, content: string, totalLength, offset, truncated?, nextOffset?}. Returns an error result with a hint to use docs_search if no match is found. The url field is the .html page, while sourceUrl is the raw .md source used for fetching guide content. Both fields are populated for script-api, guides, Salesforce Help, and tooling entries; job-step entries have neither. For those online corpora, content is fetched from sourceUrl and cached locally; offline fallback shows summary + headings + both URLs.

Examples:

• Read Script API: docs_read(query="dw.catalog.ProductMgr")
• Read guide by ID: docs_read(query="sfnext/sfnext-get-started")
• Read the next page of a long guide: docs_read(query="sfnext/sfnext-get-started", offset=12000)
• Read job step: docs_read(query="ImportCatalog")

docs_list

Enumerate documentation entries for browsing a known category. Use docs_search to answer a question; use this tool to list the contents of a category you already know. Results are a table of contents (id + title + category only) and are paginated.

Parameter	Type	Required	Default	Description
category	string	No	(directory)	Filter by category: script-api, commerce-api, pwa-kit-managed-runtime, sfnext, sfra, b2c-commerce, tooling, job-step, help-admin, help-merchant
workspace	string	No	(directory)	Workspace awareness: auto (auto-detected) or specify one or more types comma-separated: cartridges, sfra, pwa-kit-v3, storefront-next. Omit for the category directory
limit	number	No	100	Maximum entries per page
offset	number	No	0	Number of entries to skip (for pagination)

Returns (filtered): {category, total, offset, count, entries: [{id, title, category}], truncated?, nextOffset?}.

Returns (no filter): a compact category directory — {note, total, categories: [{category, count}]} — instead of the full corpus. Pass a category (or workspace) to list its entries, or use docs_search.

---

XSD Schemas

docs_schema_search

Fuzzy-search XSD schema ids.

Parameter	Type	Required	Default	Description
query	string	Yes		Schema name or partial match (e.g., "catalog", "order")
limit	number	No	20	Maximum number of results

Returns: {query, results: [{entry: {id, filePath}, score}]}.

docs_schema_read

Read the contents of an XSD schema (raw XML) plus the on-disk path.

Parameter	Type	Required	Description
query	string	Yes	Exact id or fuzzy query

Returns: {entry: {id, filePath}, content: string, path: string}. Returns an error result with a hint to use docs_schema_search if no match.

docs_schema_list

List every available XSD schema id.

No parameters.

Returns: {count, entries: [{id, filePath}]}.

---

See also

• Docs CLI commands — b2c docs search / read / schema