# Start a Workflow Source: https://docs.pitchpointsolutions.com/api/workflow/start_workflow/ **Beta** Starts a workflow with your data supplied inline, in one call. You don't start a session first: this call creates one, stages your terms into it and starts the workflow against it. This is the direct route. There are deliberately two ways to **start** a workflow — here with terms inline, or against a [staged session](https://docs.pitchpointsolutions.com/api/session/start_session_workflow/) — but only one way to **follow** one afterwards, and both routes return the same workflow envelope. Use this when you will start one workflow against a set of data. Use a session when you will start more than one against the same data, or when you build the data up over several calls: staging is free, so starting two workflows from one session sends the data once rather than twice. > **Before Starting:** Get the URL to submit against from your Pitchpoint Account Representative, along with the `service` and `model` values your account is entitled to use. > **Warning:** **This call starts a real workflow and is billed.** Each call is an independent workflow, never a retry of a previous one, unless it carries the same `Idempotency-Key` and body as an earlier call, within the key's lifetime ([Retrying Safely](https://docs.pitchpointsolutions.com/developer_guides/sessions_and_workflows/retrying_safely/)). ### Sample ```bash url="https://api.pointservices.com/riskinsight-services-ws/resources/v1/workflows" curl -X POST "${url}" \ -H "Authorization: Bearer your_access_token_here" \ -H "Content-Type: application/json" \ -H "Accept: application/json" \ -d '{ "product": { "service": "ADV", "model": "FirstOrder" }, "correlationId": "alpha_numeric_string_to_help_identify_this_workflow", "terms": [ { "person": { "firstName": "Jane", "lastName": "Sample", "ssn": "000-00-0000", "dob": "01/31/1980" } }, { "property": { "propertyUsage": "PrimaryResidence", "propertyType": "Detached", "address": { "addressLine1": "1 Sample Street", "city": "Sampleton", "state": "NY", "postalCode": "00000" } } } ] }' ``` #### Header Properties | Property | Value | Required? | |---------------------------|------------------------|-----------| | Authorization | Bearer *your token* | true | | Content-Type | application/json | true | | Accept | application/json | false | | X-PPS-UserAlias | *your value* | false | | X-PPS-UserAgent | *your value* | false | | X-PPS-CorrelationID | *your value* | false | | Idempotency-Key | *a V4 UUID* | false | Send `Content-Type: application/json` on **every** POST, including one with no body at all. In deployed environments a request that arrives without it is refused by edge protection in front of the application, with a `403` carrying an HTML body rather than any response documented here — and because the service never sees the request, nothing can explain it to you. A local or unprotected deployment may accept the same request, so do not take a working call there as evidence the header is optional. `Accept` is not enforced — a request without it is answered normally with JSON — but send it anyway, so that a future change to content negotiation cannot alter what you receive. `X-PPS-CorrelationID` is a **support-tracing header and is not the same thing** as the `correlationId` field in a request body. Only the body field is recorded against an order and searched by [Find Workflows](https://docs.pitchpointsolutions.com/api/workflow/list_workflows/); setting the header globally in your HTTP client does not make your orders findable. `Idempotency-Key` makes a retry of this call safe: send the same request again with the same key and you get the first response back, marked `X-Pps-Idempotent-Replay: true`, instead of the work being done twice. Use a new V4 UUID for each request. A key can be 1 to 255 characters; never put sensitive data in it. See [Retrying Safely](https://docs.pitchpointsolutions.com/developer_guides/sessions_and_workflows/retrying_safely/). See the [Common Headers guide](https://docs.pitchpointsolutions.com/developer_guides/headers/index/) for details on the optional `X-PPS-*` headers above. #### Request Data Properties | Property | Description | Type | |---------------|------------------------------------------------------------------------------------------------------------------------------------------|--------| | product | (mandatory) The product to run. Both `service` and `model` are required. Free text rather than a closed list; a pair that does not resolve is a 400. `model` is matched ignoring case, and a response gives it back in the spelling the product catalog uses. | object | | terms | (mandatory) The data to run against, as an array of terms. Must be present and non-empty. See [SAMI v2 Terms](https://docs.pitchpointsolutions.com/api/sami_terms/conventions/). | array | | referenceNumber | (optional) Your own identifier for this loan: the loan or case number your staff will quote when they ask about it. Recorded on the snapshot these terms create, exactly as on [Stage Data](https://docs.pitchpointsolutions.com/api/session/stage_session_sami/), and shown as the loan's reference number to Pitchpoint's fulfillment staff. Leaving it out records the snapshot without one. Leading and trailing whitespace is trimmed; longer than 255 characters is a 400 `VALIDATION_ERROR`, and nothing is started. A value that resembles SQL or markup is refused at the edge with no JSON body, so encode such a value (Base64, for instance) rather than sending it raw. Independent of `correlationId`: this API does not copy one into the other. | string | | correlationId | (optional) Your own reference for this run, at most 255 characters. Not required to be unique. This is the value [Find Workflows](https://docs.pitchpointsolutions.com/api/workflow/list_workflows/) searches. Leading and trailing whitespace is trimmed, and a blank value is treated as absent. Longer than 255 characters is a 400 `VALIDATION_ERROR`, and nothing is started. A value that resembles SQL or markup is refused at the edge with no JSON body, so encode such a value (Base64, for instance) rather than sending it raw. Not the same as a term's `correlationId`; see [Correlation ids on terms](https://docs.pitchpointsolutions.com/api/sami_terms/conventions/#correlation-ids-on-terms). | string | Fields this call doesn't define are ignored, so a misspelled field name isn't reported back to you. Five things about `terms` are checked when you submit. Each one rejects the whole request with a 400 `VALIDATION_ERROR`, rather than dropping just the bad term: 1. `terms` must be present and a **non-empty array**. `"terms": []` is rejected. 2. Every element must be a **JSON object** — never `null`, never a bare string or number. 3. Every element must carry **exactly one key** — never zero and never two. `{}` and `{"person": {…}, "property": {…}}` are both rejected. 4. That key must be **one of the eleven branch keys**: `person`, `property`, `participant`, `loanTerms`, `loanToValue`, `transactionDetail`, `loanDetail`, `refinance`, `housingExpense`, `closing`, `company`. 5. Every value must be **readable as its field's type**. A number lands in a string field and `"true"` in a boolean one, but a value such as `"yes"` for an `…Indicator` fails the whole request. Everything else is permissive. An absent field is accepted without complaint, and a field inside a term that this service does not recognize is ignored rather than refused, so a misspelled field name will not be reported back to you. Read the term descriptions as a statement of what is understood, not as a validator. ### Responses > **Warning:** **Not every response carries the same body. Branch on the `Content-Type` first, and then on what the body contains.** `application/json` is the `messages` envelope described on this page: read `messages[].code`. `application/problem+json` is a problem document, with `type`, `title`, `status`, `detail` and `instance` and no `messages` key: read `type`, and compare it as the whole URI (for example `https://pointservices.com/problems/concurrent-modification`). Some rejections carry no JSON at all. A client that calls its JSON parser on every non-2xx body, or that reads `messages[0].code` without first checking that `messages` is present, will fail on the responses it is most likely to meet. See [Error handling](https://docs.pitchpointsolutions.com/developer_guides/sessions_and_workflows/error_handling/) for the whole rule. **202** Accepted for processing. Identical in shape to the 202 on [Start a Workflow in a Session](https://docs.pitchpointsolutions.com/api/session/start_session_workflow/#responses), `sessionId` included: this route creates a session for you and stages the `terms` you sent into it. The `sessionId` it returns names an ordinary session, and is carried on every later read of the workflow. You can [read it](https://docs.pitchpointsolutions.com/api/session/get_session/), [stage](https://docs.pitchpointsolutions.com/api/session/stage_session_sami/) revised data into it, or start another workflow against it. Your `correlationId` belongs to this run only. The session this call creates does not carry it: its `session.correlationId` is one the service generated. A retry under the same `Idempotency-Key` returns this same 202, byte for byte, with the header `X-Pps-Idempotent-Replay: true`, naming the same `workflowId`. No second workflow is started or billed. See [Retrying Safely](https://docs.pitchpointsolutions.com/developer_guides/sessions_and_workflows/retrying_safely/). `workflowId` is the handle for every later read and reconfirm and is stable for the life of the workflow. `runId` identifies this one execution and **addresses nothing** — no endpoint accepts it. On a first run the two share a suffix and differ only by prefix, so use the prefix, never the suffix, to tell them apart. No timestamps are set on an acknowledgment; `ordered`, `fulfilled`, `completed` and `serviceable` all appear only on a later read. An acknowledgment is never reconfirmable; a later read's `workflow.reconfirmAvailable` tells you when it is. See [Reconfirm a Workflow](https://docs.pitchpointsolutions.com/api/workflow/reconfirm_workflow/). Example: ```json { "workflow": { "workflowId": "wfpop_0000000000002080001", "runId": "wfrunpop_0000000000002080001", "correlationId": "alpha_numeric_string_to_help_identify_this_workflow", "sessionId": "0000000000013500001", "product": { "service": "ADV", "model": "FirstOrder" }, "productType": "ADV", "productName": "FirstOrder", "status": "processing", "reconfirmAvailable": false }, "messages": [ { "code": "ACCEPTED", "text": "workflow accepted for processing", "timestamp": "2026-08-07T15:01:42Z" } ] } ``` **400** The workflow was refused and **nothing was billed**. Causes: `product` absent, malformed, or carrying only one of `service` and `model`; a pair that does not resolve to a product you may use; `terms` absent, empty or not an array, or containing an element that is not an object, carries other than exactly one key, or carries a key that is not one of the eleven branch keys; a value in a term that cannot be read as its field's type; a `correlationId` or `referenceNumber` longer than 255 characters. A literal `null` body, malformed JSON, and a `Content-Type` naming an unusable character set are refused before evaluation, as problem documents. An `Idempotency-Key` that is blank, longer than 255 characters, or already used with a different body is also refused with `VALIDATION_ERROR`. A 400 is never stored. After fixing a problem in the body, send it again with the same key. After a key-reuse 400, see [Retrying Safely](https://docs.pitchpointsolutions.com/developer_guides/sessions_and_workflows/retrying_safely/#when-to-make-a-new-key). An unresolvable product will not become resolvable on its own — correct it rather than retrying. A 400 arrives in **two different body shapes**, decided by how far the request got before it was refused. Branch on the `Content-Type`: `application/json` is the `messages` envelope, and `application/problem+json` is a problem document. **A value this service evaluated and rejected** carries the `messages` envelope, with `Content-Type: application/json` and the code `VALIDATION_ERROR`. That includes a value inside `terms` that can't be read as its field's type, such as `"yes"` for an `…Indicator`: the body as a whole was readable, so the value was evaluated. | Property | Description | Type | |----------------------|----------------------------------------------------------------------------------------------------------------|--------| | messages | (present) Always an array, never null. Carries one message whose `code` is `VALIDATION_ERROR`. | array | | messages[].code | (present) Always `VALIDATION_ERROR` on this shape. | string | | messages[].text | (present) Human-readable description of what was wrong. Never machine-parsed — the wording is not contractual. | string | | messages[].timestamp | (present) RFC 3339 UTC with a trailing `Z`. Fractional seconds appear only when they are not zero, so parse this value rather than slicing it. | string | Example: ```json { "messages": [ { "code": "VALIDATION_ERROR", "text": "the requested product does not resolve", "timestamp": "2026-08-07T15:01:42Z" } ] } ``` **A request refused before any value was evaluated** carries a problem document instead, with `Content-Type: application/problem+json` and **no `messages` key and no `VALIDATION_ERROR` code**. Its `type` tells you which case it is: | Cause | `type` | |-------|--------| | A body that could not be read as a workflow start at all: malformed JSON, a literal `null` body, a JSON array or bare value where an object was expected, or a field that can't be read as its type outside `terms` | `about:blank` | | A `Content-Type` header naming a character set this service cannot use. It is refused before the body is read, so it says nothing about what you put in it. | `https://pointservices.com/problems/invalid-request` | Example: ```json { "type": "about:blank", "title": "Bad Request", "status": 400, "detail": "Malformed JSON", "instance": "urn:pps:request:0000000000000000000000000000000000000000000000000000" } ``` > **Warning:** A client that reads `messages[0].code` on every 400 without checking that `messages` exists will throw on the problem-document shape. Branch on the `Content-Type`, or on the presence of `messages`, not on the status. No workflow is started and no handle is issued. **401** No credential the platform could accept reached the service. The body is a **JSON problem document**, not the `messages` envelope, and the response carries a `WWW-Authenticate: Bearer realm="Helix"` challenge. Read the machine-readable part from `type`; `title` and `detail` are human copy and are not contractual. | Property | Description | Type | |----------|--------------------------------------------------------------------------------------------------------|--------| | type | (present) An absolute URI identifying the problem. `about:blank` when no credential was sent at all. | string | | title | (present) Short human-readable summary. | string | | status | (present) Repeats the HTTP status code. | number | | detail | (present) Human-readable explanation. Never machine-parsed. | string | | instance | (present) An identifier for this one request. Quote it when contacting support. | string | Examples: ```json { "type": "about:blank", "title": "Unauthorized", "status": 401, "detail": "Authentication is required. Present a bearer access token in the Authorization header.", "instance": "urn:pps:request:0000000000000000000000000000000000000000000000000000" } { "type": "https://pointservices.com/problems/invalid-token", "title": "The access token was not accepted", "status": 401, "detail": "The access token was rejected.", "instance": "urn:pps:request:0000000000000000000000000000000000000000000000000000" } ``` The second form carries `WWW-Authenticate: Bearer realm="Helix", error="invalid_token"`, and is what a token the service will not accept returns. The `invalid-token` type deliberately does not say why the token was refused: the platform cannot tell an expired token from a revoked, badly signed or malformed one. Treat every `invalid-token` the same way. > **Warning:** **Refresh your token on a 401, and only on a 401.** An expiring token moves a call from 200 to 401, never to 403. Obtain a fresh access token and retry the request once. If it fails again, stop and surface the error rather than looping — repeated attempts with a credential the service has already refused will meet the rate throttle described below. **403** Most often your account is not entitled to the product requested. Check `service` and `model` against what your Account Representative provisioned. No workflow is started and nothing is billed. A 403 is an **authorization decision**: your identity was established and the answer is still no. It is never a statement about your token, so refreshing the token will not turn it into a 200. A refused request leaves nothing behind, so there is nothing to clean up. The body is a problem document (`application/problem+json`), not the `messages` envelope, so there is no `code` to branch on. Its `type` is one of two values: | `type` | Meaning | |--------|---------| | `about:blank` | The request was denied before the operation itself ran. | | `https://pointservices.com/problems/not-authorized` | A policy decision about the specific resource or product you named. | Example: ```json { "type": "https://pointservices.com/problems/not-authorized", "title": "Not authorized", "status": 403, "detail": "The authenticated identity is not permitted to perform this action.", "instance": "urn:pps:request:0000000000000000000000000000000000000000000000000000" } ``` > **Warning:** **A 403 can also come from the edge, before this service is reached,** when a request omits `Content-Type` or carries a value resembling SQL injection or cross-site scripting. That refusal carries an HTML body, not a problem document, and this service never sees it. Check the `Content-Type` of the response before you parse it. A value that legitimately contains such text should be encoded — Base64, for instance — rather than sent raw. **404** The path you called is not served. Check that the URL starts with the `/riskinsight-services-ws/resources/v1` path prefix. This path takes no identifier, so a 404 here means the path itself is not served: a wrong host, or an environment where this API is not deployed. The body is a problem document (`application/problem+json`) with `type` `about:blank`. Example: ```json { "type": "about:blank", "title": "Not Found", "status": 404, "instance": "urn:pps:request:0000000000000000000000000000000000000000000000000000" } ``` A URL missing the `/riskinsight-services-ws/resources/v1` path prefix does not reach this API at all. The 404 then comes from infrastructure in front of it, and carries no body this API defines, so don't try to parse it. If a 404 has no problem document, check the path prefix before anything else. **409** `IDEMPOTENCY_IN_PROGRESS` — an earlier request under the same `Idempotency-Key` is still running. **This is the one 409 to retry:** wait the number of seconds in the `Retry-After` header, then send the same request with the same key. If it persists for more than a few minutes, stop and follow [When the 409 doesn’t go away](https://docs.pitchpointsolutions.com/developer_guides/sessions_and_workflows/retrying_safely/#when-the-409-doesnt-go-away). A concurrent write to the same workflow was detected. This arrives as a `concurrent-modification` problem document, described below. Treat a `WORKFLOW_WRITE_CONFLICT` code in the envelope the same way. Either way the service does not retry it, so you decide whether to resubmit. **Branch on the `Content-Type` first, and only then on `messages[].code`.** A 409 here arrives in two body shapes, and the corrective action differs by cause. Never branch on the status alone, and never read `messages[0].code` without checking that `messages` is present. **`application/json`** carries the `messages` envelope: | Property | Description | Type | |----------------------|-----------------------------------------------------------------------------------------------------------|--------| | messages | (present) Always an array, never null. | array | | messages[].code | (present) One of the codes described above for this status. | string | | messages[].text | (present) Human copy. Never machine-parsed. | string | | messages[].timestamp | (present) RFC 3339 UTC with a trailing `Z`, fractional seconds only when they are not zero. | string | **`application/problem+json`** carries a problem document with `type` `https://pointservices.com/problems/concurrent-modification` and no `messages` key. Another request changed the same resource while yours was in flight, and **nothing from yours was applied**. It carries no `Retry-After`, deliberately: read the resource again, decide whether your change still applies to what you now see, and resubmit only if it does. Don't resend the same request on a timer. Example: ```json { "type": "https://pointservices.com/problems/concurrent-modification", "title": "Concurrent modification", "status": 409, "detail": "The resource was modified by another request while this one was being processed. This request was not applied; re-read the resource before resubmitting.", "instance": "urn:pps:request:0000000000000000000000000000000000000000000000000000" } ``` In either shape the conflict is final for that attempt: the service does not retry it for you. The one exception is `IDEMPOTENCY_IN_PROGRESS`, which always arrives in the `messages` envelope and which you retry after the `Retry-After` delay. **429** Refused by a rate throttle at the edge, before this service is reached. The throttle counts requests per source address across every endpoint on both the session and workflow surfaces. Because the edge sends this response before the request reaches the service, it is not described in the OpenAPI specifications or the API Explorer. The refusal carries no body, no `messages` array and no `Retry-After` header, and this service never sees it, so it will not appear in any support trace. Back off and resume. When following a workflow to completion, an interval that stays clear of the throttle is two seconds for the first two checks, then five, then ten seconds thereafter, with an overall deadline after which you stop and investigate. **500** An unhandled failure inside the service. The body is a problem document (`application/problem+json`) with `type` `about:blank`, not the `messages` envelope, so branch on the status alone. When you report the error to PPS, quote the document's `instance`: it carries the same id as the `x-pps-request-id` response header, and it is what lets PPS find this exact request. > **Warning:** **Retrying this call is not safe.** A 500 means the outcome is **unknown**, not that nothing happened: a failure raised after the work was committed leaves that work in place. **A workflow may have been started and billed.** Do not simply retry. A 500 on this call is stored: a retry under the same `Idempotency-Key` returns the same 500. Search [Find Workflows](https://docs.pitchpointsolutions.com/api/workflow/list_workflows/) for the `correlationId` you sent, and start another, under a **new** key, only if nothing comes back. Without an `Idempotency-Key`, retrying blind is how one workflow becomes two, and nothing on this surface removes a duplicate for you.