Skip to content
RUBI

Own tech

Expose these read only endpoints on your platform. Rubi searches your API during interview creation and retrieves only the selected trader and their activity.

Transport#

  • HTTPS only, TLS 1.2 or later, JSON bodies in UTF-8.
  • You choose the base URL, for example https://risk.example-firm.com/rubi/v1. Every path below is appended to it. No trailing slash.
  • Authentication is a static secret you generate, sent as Authorization: Bearer <api key>. Rotate it whenever you want and paste the new value in the Rubi dashboard. Rubi also sends Accept: application/json.
  • Restrict the API to the Rubi egress IP ranges shown under Integrations. Each range has a copy button.
  • Rubi never writes. A request without the key must not return 200.
  • Timestamps are RFC 3339 in UTC (2026-09-23T14:05:00Z). Money is a JSON number. Unknown optional fields may be null or omitted.

Pagination#

List endpoints must apply the requested scope before pagination. Never ignore a filter or fall back to returning the entire trading book.

ParameterTypeRules
searchstringRequired for interactive /traders search. At least 2 characters. Match name, email and customer ID on your server.
trader_idstringRequired on /accounts and /payouts. Every returned record must belong to this trader.
account_idstringRequired on /trades. Every returned record must belong to this account.
updated_sincetimestampSent on /trades retrieval. Inclusive lower bound on updated_at, set to 90 days before retrieval.
cursorstringOpaque value you returned as next_cursor. Absent on the first page.
limitintegerPage size. Rubi sends 8 for search and 200 for selected-trader retrieval. Accept 1 to 500. Default 100 if omitted.

The response shape is { "data": [...], "next_cursor": "..." }. next_cursor is required and must be a non-empty string ornull on the last page. Missing or malformed cursors fail the request. Never silently truncate a result set.

Order records by updated_at ascending, then id ascending, so a cursor never skips a row that was updated while a page was being read. Return each record only once within a retrieval. Repeated cursors or duplicate records fail rather than producing a partial import. Selecting the same trader again upserts their existing records. Retrieval caps are 100 pages per resource and 100,000 records per snapshot.

Filters are required
Return an error if a requested filter is unsupported. Rubi rejects data outside the selected trader or account scope and never scans the whole book as a fallback. CI or other identifiers can be searched only if your API exposes and indexes them.

Retrieve on selection#

  1. Connect the API. Rubi validates access without importing your traders or scheduling periodic polling.
  2. Type at least two characters in the interview trader search. Suggestions come from your API and remain unsaved.
  3. Select a suggestion. Rubi fetches its exact profile, that trader's accounts, their trades updated in the last 90 days and optional payouts.
  4. The selected records are saved for the interview. Selecting the same provider trader again refreshes their data.

Previously imported CRM data stays available. Legacy sync intervals are ignored; no whole-book backfill or recurring global sync runs.

Errors, limits and latency#

Use the matching HTTP status and this body:

json
{ "error": { "code": "not_found", "message": "Unknown trader cus_18422" } }
StatusWhenWhat Rubi does
401 or 403Missing or rejected key, or an IP that is not allowlisted.The request fails until you fix the key or the allowlist.
404Unknown single resource, or payouts you do not expose.A missing selected trader is an error. Payouts 404 is accepted and payouts are skipped.
429You are throttling.Rubi reads Retry-After in seconds (clamped from 1 to 60, otherwise waits 10) and retries.
5xx or timeoutYour server failed or took more than 20 seconds.Rubi retries with backoff. The dashboard shows the endpoint and the status.
200 with invalid JSONBody is not a JSON object with the expected fields.Reported as an unexpected response. The check fails.

Support at least 5 requests per second. Aim to answer within 10 seconds: the live check warns above that. The hard timeout is 20 seconds.

Connection validation#

Checks use bounded samples only. No sample is saved and the provider directory is never paged through.

  1. GET /health must return { "status": "ok" }. A request to GET /traders?limit=1 without a key must return 401 or 403.
  2. An authenticated GET /traders?limit=1 samples at most one profile. This diagnostic is the only search-filter exception.
  3. Rubi searches the exact sample ID with search and limit=5, then validates its exact GET /traders/{id} response.
  4. Up to five accounts and payouts are checked with trader_id. Up to five trades are checked with account_id from the first account.

An empty directory gets a dummy search and dependent checks are skipped. Exact diagnostic IDs may be shorter than the two-character UI minimum. Trade diagnostics may omit updated_since. Cursor pagination and ownership are checked during actual retrieval as well. The dashboard shows each check and hints for failures.

Connect from the dashboard

  1. Open Integrations and choose Own tech.
  2. Paste the base URL and the bearer key.
  3. Run the live check. Fix any failed step, then save.

The same connection can be created with POST /v1/integrations using "kind": "custom_api". See the integrations endpoints.