The `/platform/...` path *is* the complete resource path — there is **no `/connect` segment** and you do not append the operation name (details below). Use API version `67.0`.
**Send it with whatever Connect/REST API tool your environment provides** — an MCP server that invokes Connect/REST APIs, the `sf` CLI, or any REST client. Map the three primitives above to that tool's parameters.
**Whether the path needs the `/services/data/v67.0` prefix depends on the tool, not on MCP-vs-CLI** — some tools want the full versioned URL, others take the bare `/platform/...` path and add the version/host themselves. Check the tool's own docs; if a bare path returns `NOT_FOUND`, try the full versioned form (and vice-versa).
**If using the `sf` CLI** — supply the **full versioned URL** and pass the body inline. The `--body` flag takes the JSON itself, **not** a filename (a bare filename is sent verbatim → `JSON_PARSER_ERROR`):
```bash
sf api request rest "/services/data/v67.0/platform/data-resilience/archive/search" \
**If using an MCP server / other REST client** — map method, path, and body to its parameters (names vary per tool; check its docs). For example, the body for the same search is:
(For GET operations, append the inputs as query-string params on the path and send no body.)
> Two path rules that otherwise surface as a misleading error, **regardless of how you send the call**:
> - **Use the literal path from the `Method + Path` column below — never build one from the operation name.** The operation identifiers in this doc (`search-archived-records`, `unarchive-records`, …) are *names*, not URL segments. The real path is short and rarely matches the name: `searchArchivedRecords` → `POST /search` (NOT `/search-archived-records`), `unarchiveRecords` → `POST /unarchive`, `getArchiveStorageUsed` → `GET /storage/archive-used`. Synthesizing the path from the operation name (e.g. kebab-casing it and appending it) produces a route that returns **404** even though the operation exists. Always copy the path string verbatim from the table.
> - **The path stops at `/platform/data-resilience/archive/...`** — there is no `/connect` segment, even though this is a Connect API. Inserting one returns `NOT_FOUND` for the whole namespace, which can read as "Archive is disabled" — it isn't; it's the path.
The first column is the **operation name** used to refer to each op in this skill — it is **not** part of the path. The method and path to send are in the **Method + Path** column; each path shown (e.g. `POST /search`) is shorthand for `POST /platform/data-resilience/archive/search`.
| Operation (name only — NOT a path segment) | Purpose | Method + Path (send THIS) | Verify with |
Search archived records by object, filters, date ranges, and sort.
**Required**: `sobjectName` + at least 1 filter. Missing them returns a clean envelope validation error (`statusCode 400`).
**Inputs**:
-`sobjectName`*(string)* — API name of the sObject to search.
-`filters`*(array)* — Filter conditions, each `{field, value}` where **both are required strings**. `value` is a single string — **not** an array, **not** nullable (null/omitted → `400 "This field may not be null"`). There is **no `operator` field**. **At least 1, up to 6 filters, combined with AND only** (OR is not supported). Example: `[{"field":"Subject","value":"Foo"}]`.
-`dateRanges`*(array)* — Primary date filter: array of `{field, from, to}`. `from`/`to` must be full ISO-8601 datetime (`"2020-01-01T00:00:00Z"`); date-only → `400 JSON_PARSER_ERROR` (xsd:dateTime). Use the special field `archive_date` to filter by archive date instead of `CreatedDate`/`ModifiedDate`.
-`dateRange`*(object)* — Optional **singular** convenience range `{field, from, to}`; treated as a one-element `dateRanges`. Same singular shape that `unarchive-records` uses.
**Output**: HTTP **201**, `body = { records[], total_result_count, scroll_id }`, `body.statusCode = 200`, `errorMessage` null on success. **Branch on `body.statusCode`, not the HTTP code** — once past request validation the response is always 200/201 even on logical failure, with the error in `errorMessage` + `statusCode`.
**Excluded objects**: `Feed`, `History`, `Relation`, `Share` are not searchable; Files/Attachments are not retrievable.
### Pagination
Read records inline from each response. If `body.scroll_id != "-1"`, call `get-search-archived-records-next-page` with that `scroll_id`. **STOP when `scroll_id == "-1"`** — calling next-page with `"-1"` (the terminal sentinel) → 500. There is no separate fetch-by-requestId step.
Archive search that enforces the user's sharing rules. (There is no `/search/related` endpoint — this is the operation that takes the JSON filter map.) Gated by the **`ViewArchivedRecords`** user permission (unlike plain `search-archived-records`, which uses `ViewSearchPage`).
-`objectName`*(string)* — API name of the sObject.
-`filtersJson`*(string)* — JSON-encoded **OBJECT MAP** of `fieldName → value`, e.g. `"{\"Subject\":\"Foo\",\"Status\":\"New\"}"`. **NOT** an array of `{field,value}` objects — the array form is rejected with `isSuccess:false "No valid filters provided. Please provide filtersJson"`.
-`dateField`*(string)* — API name of the date field for temporal filtering.
`run-analyzer` triggers the analyzer; HTTP 201, output `message`*(string, human-readable status)*. **`isRunning` is ALWAYS `null`** — only `message` is populated; never branch on `isRunning`. Poll `get-analyzer-report` to confirm completion. Non-destructive / idempotent.
-`dateRange`*(object)* — optional **SINGULAR** range `{field, from, to}` (unlike `/search`, which uses the plural `dateRanges`); full ISO-8601 datetimes. Omit to unarchive by filters alone.
**Caps**: ≤1000 matched records (else not processed); ≤50 unarchive requests/hour/org. Restores the **whole archived hierarchy** of each match. Requires the **`UnarchiveSdk`** user permission (on top of org-level Archive enablement).
**Output**: `body` (unarchive job details incl. job id), `errorMessage`, `statusCode`. **Verify** by re-running `search-archived-records`. **Rollback**: re-archive via a new archive job with the same criteria.
**Input**: `criteria`*(array of `{sobject, field, value}`)* — ≤10 items, **one per object type**; ≤10,000 root records erased per org/day; field/object names case-insensitive. Deletes the **entire archived hierarchy** of each match (no partial deletion). Note: the criterion must match a record archived **as a root** — filtering a parent (e.g. Account by Id) when only its children were archived matches nothing.
**Output**: HTTP 201, `body.request_id`*(UUID)*. Poll `get-rtbf-status` (path param `requestId` = that UUID) → `body.status` (e.g. `"Request is open. Scan is still in progress"`). **Rollback**: none — RTBF erasure is permanent.
Submits a PII-masking (anonymization) request — irreversibly replaces detected PII values with placeholders (e.g. `redacted@example.com`) while keeping the record + non-PII fields searchable.
**Input**: `criteria`*(array of `{sobject, field, value}`)* — same shape as RTBF.
**Behavior**: permanent; one-time per record (re-requests on an already-masked record are ignored); shares the 10,000/day RTBF rate limit; **PII fields are auto-detected** (you cannot choose them); records under legal hold / retention lock are excluded; cascades to child records. Available **only via this API** (not in the Archive UI).
Both are `GET /log/...` and mint a one-time presigned download URL. `get-execution-details-stream-url` → the execution-detail log; `get-failed-records-stream-url` → the failed-records log (records that did not process). Identical contract.
**Output**: `{ url }`. **`url != null` is success; `url: null` means no log was resolved** (missing/incorrect `requestId`/`reportType`, or the activity produced no such log) — always check `url != null`; never treat `url:null` or the 201 status alone as success.
**Required inputs**:
-`requestId`*(string)* — the **`ArchiveActivity` Id** (`8qv…` key) of a completed, log-producing job present in the archiver backend — **not** a search requestId. A missing/non-matching id → `url:null`.
-`reportType`*(string)* — that activity's `Type`: `Archive | Unarchive | Analyzer | Purge | Export-to-external-bucket | Export-and-download`. Omitting it → `url:null`.
-`sobjectName`*(string, optional)* — the backend self-resolves it.
This is the bridge between `ArchiveActivity` (see `archive-activity-entity.md`) and downloadable logs.
Returns `body.usedStorage[4]` (doubles — per-tier bytes consumed) and `body.availableStorage[4]` (per-tier capacity) — two **parallel positional arrays** (NOT a flat metric, NOT key/value maps). The 4 slots are the same in both:
**`availableStorage[2]` and `[3]` (the archive tier) are ALWAYS 0**: archive storage is unmetered, so a `0` there means "not tracked", NOT "no space left". Only `availableStorage[0]`/`[1]` (live org data/file remaining) are real. All values rounded to 2 decimals.
