This changelog is organized in reverse chronological order, with the most recent updates appearing first. Back to our documentation for futher API details.
The changelog updates initiated on 2024-10-25.
AWAITING_TRAVEL_RULE_REVIEW may go back to AWAITING_TRAVEL_RULE. Its travel_rule is null again: send the declaration again, in full, for the same operation. The declaration sent again is always reviewed, whatever the answers, the purpose or the amount: it goes to AWAITING_TRAVEL_RULE_REVIEW, after the documents when the purpose asks for them, and the operation is never credited or released directly. A withdrawal back in this state has no deadline and is never canceled automaticallyOCO (One-Cancels-the-Other)POST /rest/v3/orders accepts the new type value OCO. It places a take-profit LIMIT at price and arms a stop-loss STOP_LIMIT that triggers at stop_price and is placed at the new field stop_limit_price, both under a single order id and a single client_order_id. The first leg to activate wins and the other is canceled: a first trade of the take-profit cancels the stop-loss, and the trigger of the stop-loss cancels the take-profit. POST /rest/v3/orders/batch and POST /rest/v3/orders/cancel-replace take it as wellquantity, price, stop_price and stop_limit_price are required. On a SELL, price must be above stop_price and stop_limit_price at or below it; on a BUY, price must be below stop_price and stop_limit_price at or above it. A body that breaks the rule is answered with 400 Bad Request naming it: on a SELL OCO, price must be above stop_price and stop_limit_price must be at or below stop_price, or the same for a BUY, mirroredpost_only and time_in_force are refused on an OCO with 400 Bad Request, whatever their value, false and GTC included (post_only is not accepted on OCO orders): an OCO is always GTC and never post-only. stop_limit_price is refused on every other order type (stop_limit_price is only accepted on OCO orders)price and stop_price, and price and stop_limit_price to fit the market's price limits. An order that does not is refused with 422, code 4008, or code 5005 for a price outside the market's price band. The 4008 detail is generic and does not say which rule failedBUY OCO locks max(price, stop_limit_price) × quantity of the quote currency, since either leg may execute; a SELL locks quantity of the base currency, as a limit order doesprice changes once, to stop_limit_price. stop_price holds the stop-loss trigger throughoutGET /rest/v3/orders, GET /rest/v3/orders/{order_sn}, GET /rest/v3/orders/by-order-id/{id} and GET /rest/v3/orders/by-client-order-id/{client_order_id} return two new fields. legs lists the take-profit and the stop-loss, in that order, each with its role, type, price, stop_price, state and cancellation_reason. oco_winner_leg is the leg the order committed to: STOP_LOSS once stop_price triggers, TAKE_PROFIT after a first trade, null while both legs are active. Both fields are null on every other order typecancellation_reason gains the code 109: an OCO canceled because its stop price was already reached when it reached the matching queue, or because the market's matching engine does not support OCO ordersOCO in order_type on GET /rest/v3/markets; on any other market the order is refused with 422, code 4008STOP_LIMIT — a price on the wrong side of stop_price is refusedPOST /rest/v3/orders refuses a STOP_LIMIT whose price sits on the wrong side of stop_price — above it on a SELL, below it on a BUY — because such an order would trigger and never execute. It is answered with 400 Bad Request naming the rule: on a SELL STOP_LIMIT, price must be at or below stop_price or on a BUY STOP_LIMIT, price must be at or above stop_price. A price equal to stop_price is acceptedPOST /rest/v3/orders/batch the whole batch is refused, its valid orders included; in POST /rest/v3/orders/cancel-replace the order to cancel is left untouchedOCO (One-Cancels-the-Other)POST /rest/v3/orders accepts the new type value OCO. It places a take-profit LIMIT at price and arms a stop-loss STOP_LIMIT that triggers at stop_price and is placed at the new field stop_limit_price, both under a single order id and a single client_order_id. The first leg to activate wins and the other is canceled: a first trade of the take-profit cancels the stop-loss, and the trigger of the stop-loss cancels the take-profit. POST /rest/v3/orders/batch and POST /rest/v3/orders/cancel-replace take it as wellquantity, price, stop_price and stop_limit_price are required. On a SELL, price must be above stop_price and stop_limit_price at or below it; on a BUY, price must be below stop_price and stop_limit_price at or above it. A body that breaks the rule is answered with 400 Bad Request naming it: on a SELL OCO, price must be above stop_price and stop_limit_price must be at or below stop_price, or the same for a BUY, mirroredpost_only and time_in_force are refused on an OCO with 400 Bad Request, whatever their value, false and GTC included (post_only is not accepted on OCO orders): an OCO is always GTC and never post-only. stop_limit_price is refused on every other order type (stop_limit_price is only accepted on OCO orders)price and stop_price, and price and stop_limit_price to fit the market's price limits. An order that does not is refused with 422, code 4008, or code 5005 for a price outside the market's price band. The 4008 detail is generic and does not say which rule failedBUY OCO locks max(price, stop_limit_price) × quantity of the quote currency, since either leg may execute; a SELL locks quantity of the base currency, as a limit order doesprice changes once, to stop_limit_price. stop_price holds the stop-loss trigger throughoutGET /rest/v3/orders, GET /rest/v3/orders/{order_sn}, GET /rest/v3/orders/by-order-id/{id} and GET /rest/v3/orders/by-client-order-id/{client_order_id} return two new fields. legs lists the take-profit and the stop-loss, in that order, each with its role, type, price, stop_price, state and cancellation_reason. oco_winner_leg is the leg the order committed to: STOP_LOSS once stop_price triggers, TAKE_PROFIT after a first trade, null while both legs are active. Both fields are null on every other order typecancellation_reason gains the code 109: an OCO canceled because its stop price was already reached when it reached the matching queue, or because the market's matching engine does not support OCO ordersOCO in order_type on GET /rest/v3/markets; on any other market the order is refused with 422, code 4008STOP_LIMIT — a price on the wrong side of stop_price is refusedPOST /rest/v3/orders refuses a STOP_LIMIT whose price sits on the wrong side of stop_price — above it on a SELL, below it on a BUY — because such an order would trigger and never execute. It is answered with 400 Bad Request naming the rule: on a SELL STOP_LIMIT, price must be at or below stop_price or on a BUY STOP_LIMIT, price must be at or above stop_price. A price equal to stop_price is acceptedPOST /rest/v3/orders/batch the whole batch is refused, its valid orders included; in POST /rest/v3/orders/cancel-replace the order to cancel is left untouchedEvery crypto withdrawal of the public API is declared through POST /rest/v3/withdrawals/{withdrawal_sn}/travel_rule, so the documentation and the withdrawal read no longer describe the previous flow.
GET /rest/v3/withdrawals/{withdrawal_sn} no longer returns travel_rule_version. It only told the declaration flow apart from the previous one, which no withdrawal of the public API follows anymore. travel_rule is still omitted for fiat and null when no declaration is recordedGET /rest/v3/withdrawals and GET /rest/v3/withdrawals/{withdrawal_sn} no longer document the deprecated AWAITING_CONFIRMATION state, and neither do the answers of the withdrawal declaration and documents routestravel_rule_version precondition, the 422 for a withdrawal on that flow and the effective-date notice saying withdrawals keep following it are gone from themPOST /rest/v3/deposits/{deposit_sn}/travel_rule and POST /rest/v3/withdrawals/{withdrawal_sn}/travel_rule document the 400 and 422 body, { "error": { "message", "code", "details" } }, with one example per case. details lists every problem at once; code is 400 for a body that does not fit the path, 422 for an operation that cannot take the declaration and 4001 when the declaration itself is refused (a field the account type does not take, an invalid CPF/CNPJ, a purpose the operation does not accept, or purpose_description missing above USD 50,000)travel_rule of POST /rest/v3/withdrawals, and its echo in the creation response, are no longer documented. They are still accepted; declare a held withdrawal through POST /rest/v3/withdrawals/{withdrawal_sn}/travel_ruleGET /rest/v3/travel_rule/operation_reasons is no longer documented: it only served the inline travel_rule. A declaration takes the purposes of its own operation, from GET /rest/v3/deposits/{deposit_sn}/travel_rule/operation_reasons or GET /rest/v3/withdrawals/{withdrawal_sn}/travel_rule/operation_reasonsPOST /rest/v3/deposits/{deposit_sn}/travel_rule and POST /rest/v3/withdrawals/{withdrawal_sn}/travel_rule no longer take answers[]. The body is one of four variants, picked by wallet_owner (SELF or THIRD_PARTY) and wallet_type (SELF_HOSTED or HOSTED): own private wallet, own Exchange/VASP account, third-party private wallet and third-party Exchange/VASP account. The counterparty goes under sender on a deposit and recipient on a withdrawal, with the same names and shape the reads return. Each variant documents its fields and when each one is askedBR for Brazil: vasp_entity_country, sender.residence_country / recipient.residence_country and the client's fx_residence_country (BR when omitted). Countries come from GET /travel_rule/countries and the BACEN code from the operation reasons endpoint of the operation, sent as operation_reasonentity_category and same_economic_group — are required for a company account on the paths that ask them and refused for a personal account (422); institution_type is required (400) when entity_category is FINANCIAL_INSTITUTION400 Bad Request naming the field when one the path asks is missing, one is not part of the variant or a value is malformed (a field of the variant the path does not ask is ignored); 404 when the deposit or withdrawal is not yours; 422 when it is not held for the Travel Rule, was already declared, a field the account type requires is missing or one it does not accept was sent, or a CPF/CNPJ has an invalid check digittravel_rule of GET /rest/v3/deposits, GET /rest/v3/deposits/{deposit_sn} and GET /rest/v3/withdrawals/{withdrawal_sn}, and the declaration responses): client_fx_residence_country becomes fx_residence_country; economic_group_link (40/50) becomes same_economic_group, a boolean that is null when the question was not asked; the client's category moves from sender.entity_category / recipient.entity_category to the top-level entity_category and institution_type, and the ones under sender / recipient keep only the counterparty'srecipient.tax_id_number, as the reads and the deposit already didwallet_owner SELF), sender / recipient carry the client's own name, tax_id_number and document_typePOST /rest/v3/deposits/{deposit_sn}/travel_rule and POST /rest/v3/withdrawals/{withdrawal_sn}/travel_rule) and the reads (travel_rule of GET /rest/v3/deposits, GET /rest/v3/deposits/{deposit_sn} and GET /rest/v3/withdrawals/{withdrawal_sn}) now return travel_rule.required_documents: the documents the operation still waits for, with title, description and requirements, empty when it waits for none. The operation reasons endpoints only answer before the declaration; this is where the files to send are listed after itreview_required. An operation under review reads state AWAITING_TRAVEL_RULE_REVIEWGET /rest/v3/deposits/{deposit_sn}/travel_rule/operation_reasons and GET /rest/v3/withdrawals/{withdrawal_sn}/travel_rule/operation_reasons no longer return requires_approvalPOST /rest/v3/deposits/{deposit_sn}/travel_rule now takes the Travel Rule declaration of a crypto deposit held in AWAITING_TRAVEL_RULE; since 2026-10-21 its body is one of four variants (see that entry). The previous originator payload (wallet_owner, wallet_type, country, operation_reason and the originator identity) is no longer accepted and answers 400 Bad RequestGET /rest/v3/deposits/{deposit_sn}/travel_rule/operation_reasons: the BACEN operation reasons (purposes) the declaration of that deposit accepts, localized by the Locale header. Each entry tells whether the purpose requires documents and, through required_documents, which documentsACCEPTED), held for documents (documents_required, sent through the documents endpoint below) or under review (new state AWAITING_TRAVEL_RULE_REVIEW), after which the deposit is credited or REJECTED. The declaration itself never refuses the deposit201 Created with the deposit state, the outcome flags and the persisted declaration, in the same shape the deposit reads return; 400 Bad Request on an invalid body; 404 when the deposit is not yours; 422 when the deposit is not held for the Travel Rule, was already declared or an answer is missing or invalidPOST /rest/v3/deposits/{deposit_sn}/travel_rule/documents (multipart/form-data) for the purposes that require documents: send every file in a single request, signed with X-FB-CONTENT-SHA256 over the concatenated file bytes. Accepts pdf, jpg, jpeg, png, csv, xls, xlsx, up to 20 files and 25MB. The deposit is checked before any file is read (yours, held in AWAITING_TRAVEL_RULE and still owing the documents of its declaration); it then goes to review or is creditedtravel_rule of GET /rest/v3/deposits and GET /rest/v3/deposits/{deposit_sn} now returns the Travel Rule declaration, the same shape as the withdrawal one. The sender is grouped under sender (name, tax_id_number, document_type, person_type, residence_country, entity_category, institution_type), next to vasp_entity_country, client_fx_residence_country, economic_group_link, purpose_description, documents_required and the review; country is removed. sender.tax_id_number is always masked, the declaration response included. Deposits declared on the previous questionnaire come in the same shape, with the newer fields nullGET /rest/v3/deposits and GET /rest/v3/deposits/{deposit_sn} document the new state AWAITING_TRAVEL_RULE_REVIEW. Which Travel Rule step a deposit waits for is read from state, with travel_rule.documents_required and travel_rule.reviewGET /rest/v3/withdrawals/{withdrawal_sn} (travel_rule) and by POST /rest/v3/withdrawals/{withdrawal_sn}/travel_rule (travel_rule) now groups the recipient under recipient. recipient_name, recipient_tax_id_number, recipient_document_type, recipient_person_type, recipient_residence_country, recipient_entity_category and recipient_institution_type become recipient.name, recipient.tax_id_number, recipient.document_type, recipient.person_type, recipient.residence_country, recipient.entity_category and recipient.institution_type. The values do not change, and recipient is always present, with null in the fields the path did not asksenderTRAILING_STOP_MARKETPOST /rest/v3/orders accepts TRAILING_STOP_MARKET: a stop whose trigger follows the market in the order's favour and never moves back. A sell trails the high and its stop only rises; a buy trails the low and its stop only falls. When the market turns and reaches the stop, the order becomes a market orderstop_price is where the trigger starts. It is not where the order fires: from the first trade on, the trail moves it, and the value you read on the order is the one in force. It has to sit inside the market's price band — a range around the last traded price, set per market — so a trailing stop cannot be armed at a distance the book has no relation to. On every market that offers trailing stop orders the factor is 5: the band runs from the last traded price divided by 5 to the last traded price multiplied by 5. A value outside it is answered with price_out_of_range (code 5005)trail_type says how the distance is read. AMOUNT is an absolute value in the quote currency; PERCENT is a whole percentage of the price the stop trailstrail_value is the distance itself. With AMOUNT it has to be a multiple of the market's price_increment from GET /markets — the smallest difference between two valid prices. A market whose increment is 1 takes 1, 2, 500; one whose increment is 0.01 takes 0.01, 0.5, 2.55. A distance finer than the increment is rejected rather than rounded, because the engine truncates the stop to the market's precision and the stop would move by a different amount than the one you sent. It is also capped against the book: the ceiling is a multiple of the last traded price, set per market, and is 5 times the last traded price on every market that offers trailing stop orders — the same factor that bounds stop_price. A value above it is answered with price_out_of_range (code 5005). With PERCENT it is a whole number above 0 and below 100: 1, 2, 50. A fraction such as 2.5 is rejectedstop_price 400000 and trail_value 5000, keeps the stop 5,000 below the highest price traded. The market rises to 420000 and the stop follows to 415000; the market dips and the stop stays where it is. The order triggers once a trade prints at 415000 or below — protecting the rise it capturedprice is rejected: the order has none, and a triggered TRAILING_STOP_MARKET takes the book as it finds itstop_price (the trigger in force), origin_stop_price (the one sent at creation), stop_price_updated_at (null until the trail moves for the first time), trail_type and trail_value. Comparing the first two shows how far the trail has comestate: ACTIVE. There is no event of its own: what identifies it is stop_price changing while quantity_executed stays at zeroorder_type on GET /markets. Up to 5 open trailing stop orders per accountTRAILING_STOP_LIMITPOST /rest/v3/orders accepts TRAILING_STOP_LIMIT: the same trailing trigger as above, except that it becomes a limit order instead of a market one, so the fill price is boundedlimit_offset is required and says how far the limit price sits from the stop price when the order triggers. A sell subtracts it, a buy adds it. Like trail_value with AMOUNT, it is a price distance and has to be a multiple of the market's price_incrementprice is rejected at creation and stays null while the order rests: it is derived at trigger time from the stop price the trail arrived at, and only then written to the orderstop_price, trail_type, trail_value, the response fields, the WebSocket behaviour and the limit of 5 open trailing stop orders per account — works as described for TRAILING_STOP_MARKETTRAILING_STOP_MARKETPOST /rest/v3/orders accepts TRAILING_STOP_MARKET: a stop whose trigger follows the market in the order's favour and never moves back. A sell trails the high and its stop only rises; a buy trails the low and its stop only falls. When the market turns and reaches the stop, the order becomes a market orderstop_price is where the trigger starts. It is not where the order fires: from the first trade on, the trail moves it, and the value you read on the order is the one in force. It has to sit inside the market's price band — a range around the last traded price, set per market — so a trailing stop cannot be armed at a distance the book has no relation to. On every market that offers trailing stop orders the factor is 5: the band runs from the last traded price divided by 5 to the last traded price multiplied by 5. A value outside it is answered with price_out_of_range (code 5005)trail_type says how the distance is read. AMOUNT is an absolute value in the quote currency; PERCENT is a whole percentage of the price the stop trailstrail_value is the distance itself. With AMOUNT it has to be a multiple of the market's price_increment from GET /markets — the smallest difference between two valid prices. A market whose increment is 1 takes 1, 2, 500; one whose increment is 0.01 takes 0.01, 0.5, 2.55. A distance finer than the increment is rejected rather than rounded, because the engine truncates the stop to the market's precision and the stop would move by a different amount than the one you sent. It is also capped against the book: the ceiling is a multiple of the last traded price, set per market, and is 5 times the last traded price on every market that offers trailing stop orders — the same factor that bounds stop_price. A value above it is answered with price_out_of_range (code 5005). With PERCENT it is a whole number above 0 and below 100: 1, 2, 50. A fraction such as 2.5 is rejectedstop_price 400000 and trail_value 5000, keeps the stop 5,000 below the highest price traded. The market rises to 420000 and the stop follows to 415000; the market dips and the stop stays where it is. The order triggers once a trade prints at 415000 or below — protecting the rise it capturedprice is rejected: the order has none, and a triggered TRAILING_STOP_MARKET takes the book as it finds itstop_price (the trigger in force), origin_stop_price (the one sent at creation), stop_price_updated_at (null until the trail moves for the first time), trail_type and trail_value. Comparing the first two shows how far the trail has comestate: ACTIVE. There is no event of its own: what identifies it is stop_price changing while quantity_executed stays at zeroorder_type on GET /markets. Up to 5 open trailing stop orders per accountTRAILING_STOP_LIMITPOST /rest/v3/orders accepts TRAILING_STOP_LIMIT: the same trailing trigger as above, except that it becomes a limit order instead of a market one, so the fill price is boundedlimit_offset is required and says how far the limit price sits from the stop price when the order triggers. A sell subtracts it, a buy adds it. Like trail_value with AMOUNT, it is a price distance and has to be a multiple of the market's price_incrementprice is rejected at creation and stays null while the order rests: it is derived at trigger time from the stop price the trail arrived at, and only then written to the orderstop_price, trail_type, trail_value, the response fields, the WebSocket behaviour and the limit of 5 open trailing stop orders per account — works as described for TRAILING_STOP_MARKETAWAITING_TRAVEL_RULE; a withdrawal created with an auto-approval API key is confirmed by the request itself and enters it right after creationPOST /rest/v3/withdrawals/{withdrawal_sn}/travel_rule so members submit the Travel Rule declaration for a crypto withdrawal held in AWAITING_TRAVEL_RULEanswers[], one { question_key, value } per question the path asks. The question keys, options and visibility rules are documented on the endpoint; country answers come from GET /travel_rule/countries and the BACEN code from the operation reasons endpoint belowGET /rest/v3/withdrawals/{withdrawal_sn}/travel_rule/operation_reasons: the BACEN operation reasons (purposes) the declaration of that withdrawal accepts, localized by the Locale header. Each entry tells whether the purpose requires documents and, through required_documents, which documents, so the set is known before the declaration is submittedPROCESSING), held for documents (documents_required, sent through the documents endpoint below), under review (new state AWAITING_TRAVEL_RULE_REVIEW) or rejected with the funds returned (new rejection reasons TRAVEL_RULE_COUNTRY_RESTRICTED, TRAVEL_RULE_LIMIT_EXCEEDED and TRAVEL_RULE_REJECTED after review). A withdrawal held for the declaration or for documents is canceled, with the funds returned, when 24 hours pass without themGET /rest/v3/withdrawals and GET /rest/v3/withdrawals/{withdrawal_sn} document two new states: AWAITING_TRAVEL_RULE_REVIEW, described above, and AWAITING_CONFIRMATION, deprecated: a withdrawal declared before the member confirmed it, waiting for that confirmation201 Created with the withdrawal state, the outcome flags and the persisted travel_rule declaration; 400 Bad Request on an invalid body; 422 on validation errors (withdrawal not found or not awaiting the Travel Rule, missing or invalid answer) and when the declaration itself rejects the withdrawal, which ends REJECTED with the funds returnedPOST /rest/v3/withdrawals/{withdrawal_sn}/travel_rule/documents (multipart/form-data) for the purposes that require documents: send every file in a single request, signed with X-FB-CONTENT-SHA256 over the concatenated file bytes. Accepts pdf, jpg, jpeg, png, csv, xls, xlsx, up to 20 files and 25MB. The withdrawal is checked before any file is read (yours, held in AWAITING_TRAVEL_RULE for the declaration); it then goes to review or is releasedGET /rest/v3/withdrawals/{withdrawal_sn} now returns the travel_rule declaration of crypto withdrawals on travel_rule_version V2 (null otherwise or while none was submitted), including the review state; recipient_tax_id_number is masked on read. It also returns travel_rule_version: V2 for the declaration flow, V1 (deprecated) for the inline travel_ruleGET /rest/v3/travel_rule/countries listing the ISO 3166-1 alpha-2 countries accepted in the declaration, localized by the Locale header and cached for 10 minutesGET /rest/v3/travel_rule/operation_reasons now exposes requires_documents, requires_approval and required_documents per entry. It keeps the BRL filters for deposits and for the deprecated inline travel_rule.operation_reason of POST /withdrawals; the purposes of a withdrawal declaration come from the withdrawal's own operation reasons endpointtravel_rule object of POST /rest/v3/withdrawals is deprecated. Create the withdrawal without it and submit the declaration through the new endpoint while the withdrawal is in AWAITING_TRAVEL_RULEAWAITING_TRAVEL_RULE for the new endpointGET /orders/by-client-order-id finds ids above 2^53GET /orders/by-client-order-id/{client_order_id} rounded an id above 9007199254740991 (2^53 - 1) before searching, so it returned the wrong order or none. It now looks up the exact value0 to 9223372036854775807 (2^63 - 1), sent as digits only. Anything else, including a negative number, is answered with 400 and the detail "client_order_id must be an integer between 0 and 9223372036854775807"client_order_id validation on order creation and cancellation. POST /orders, POST /orders/batch, PUT /orders/cancel (by CLIENT_ORDER_ID) and POST /orders/cancel-replace now accept client_order_id only as an integer from 0 to 9223372036854775807 (2^63 - 1), sent as digits. Anything else is answered with 400 and the same detail as above. Some of these values used to be accepted: a sign (+1 was stored as 1, -1 was kept as is) and values above 18446744073709551615 (2^64 - 1), which were kept as sent. If your integration uses any of these forms, move to digits within the range before this change reaches youclient_order_id is now 2^63 - 1, the largest value an order keepsGET /orders/by-client-order-id finds ids above 2^53GET /orders/by-client-order-id/{client_order_id} rounded an id above 9007199254740991 (2^53 - 1) before searching, so it returned the wrong order or none. It now looks up the exact value0 to 9223372036854775807 (2^63 - 1), sent as digits only. Anything else, including a negative number, is answered with 400 and the detail "client_order_id must be an integer between 0 and 9223372036854775807"client_order_id validation on order creation and cancellation. POST /orders, POST /orders/batch, PUT /orders/cancel (by CLIENT_ORDER_ID) and POST /orders/cancel-replace now accept client_order_id only as an integer from 0 to 9223372036854775807 (2^63 - 1), sent as digits. Anything else is answered with 400 and the same detail as above. Some of these values used to be accepted: a sign (+1 was stored as 1, -1 was kept as is) and values above 18446744073709551615 (2^64 - 1), which were kept as sent. If your integration uses any of these forms, move to digits within the range before this change reaches youclient_order_id is now 2^63 - 1, the largest value an order keeps401 instead of the 500 "An unknown error occurred while processing the request." it used to collapse into. This covers an access key that does not exist, is disabled or has expired, and an access token that is invalid or disabled. These keep the codes they already had: 2003 "Invalid access key." for the key cases and 2003 "Invalid session." for the token ones401 and the new error code 2006 ("Timestamp out of window."). The response names the clock, so a client drifting from server time can act on it. The fix for it is to synchronize the machine clock, not to retry or to rotate the API key2006 to your authentication handling. The other cases above already used codes your client knows; what changed is the status, from 500 to 4012006 is distinct from 2004 ("Invalid timestamp."), which still means the X-FB-ACCESS-TIMESTAMP header is missing or unparseable, and from 4014 ("Timestamp is out of range."), which still means the request fell outside the X-FB-RECEIVE-WINDOW you asked forPOST /orders/batch and POST /orders/cancel-replace now return id as a string, matching POST /orders, PUT /orders/cancel and the documented order GET responses, which already did. If you parse this field as a number, update your client: order ids are numeric values transported as strings so they survive above 2^53order_id in GET /trades and to cancellation_reason_order_id in the order responsesnull are declared as nullable: price, instant_amount, instant_amount_executed, remark, cancellation_reason, client_order_id in orders, and category, deposit_info, withdraw_info and the withdrawal fees in currenciesbids and asks in GET /markets/{market_symbol}/orderbook are documented as arrays of [price, quantity] pairs, and candlesticks as arrays of fixed-length tuples, with the meaning of each index described in the endpointGET /markets/quotes documents amount and quantity as query parameters. Provide one of them; when both are sent, amount takes precedencelast_trade.date is optional: markets with no trade do not return itPUT /orders/cancel and POST /orders/cancel-replace reject a non-numeric id with 400 instead of coercing it. A numeric id keeps workingGET /markets/quotes answers 400 instead of 500 when a malformed amount or quantity is sentX-FB-RATE-LIMIT-REQUESTS-LIMIT (maximum requests in the current window, already reflecting any account-specific increase), X-FB-RATE-LIMIT-REQUESTS-REMAINING (requests still available) and X-FB-RATE-LIMIT-REQUESTS-RESET (seconds until the window resets)429 Too Many Requests responses now also include the X-FB-RATE-LIMIT-RETRY-AFTER header with the number of seconds to wait before retryingPOST /rest/v3/prime_desk/quote/request now accepts amount (the quote-currency value to trade) as an alternative to quantity (base currency) — send one or the other, never both{ "market_symbol": "btcbrl", "side": "SELL", "amount": "1000" } quotes the sale of R$1000 worth of BTC; the response returns the derived quantity and the locked pricePOST /rest/v3/deposits/{deposit_sn}/travel_rule so members submit Travel Rule originator information for a crypto deposit held awaiting travel rule, releasing itwallet_owner (SELF | THIRD_PARTY), wallet_type (SELF_HOSTED | HOSTED), country (ISO 3166-1 alpha-2), operation_reason are requiredname, tax_id_number, document_type (CPF | CNPJ | TIN), vasp_name — required combinations match the crypto withdrawal Travel Rule contractwallet_owner is SELF, name and tax_id_number are filled automatically from the member and do not need to be sent201 Created with the persisted travel_rule object (enum values UPPERCASE) on success; 400 Bad Request on an invalid Travel Rule body; 422 on validation errors (e.g. deposit not found, deposit not awaiting travel rule, or unknown BACEN operation reason)travel_rule object to the request and response body of the crypto withdrawal endpoint, allowing partners to attach beneficiary information required by the Travel Rule regulationPOST /rest/v3/withdrawalswallet_owner (SELF | THIRD_PARTY), wallet_type (SELF_HOSTED | HOSTED), country (ISO 3166-1 alpha-2), operation_reason, plus name, tax_id_number, document_type (CPF | CNPJ | TIN) or vasp_name depending on the variantwallet_owner, wallet_type) combination; see the request body variantstravel_rule is only accepted on crypto withdrawals; on the response, the field is omitted on fiat and on crypto when no information was registeredGET /travel_rule/operation_reasonsamount (BRL value) or quantity + currency_symbol (crypto quantity + currency symbol)Added support for Ed25519 API Keys (asymmetric key pair model). Clients generate the key pair locally and register only the public key with Foxbit. The private key never leaves the client environment.
The HMAC model remains fully supported and unchanged.
Authentication headers (X-FB-ACCESS-KEY, X-FB-ACCESS-TIMESTAMP, X-FB-ACCESS-SIGNATURE) and the stringToSign format are identical for both models.
The signature is hex-encoded (128 characters for Ed25519, 64 characters for HMAC).
See the Signing a request with asymmetric keys section under Authentication for key generation steps, a Node.js signing example and the headers used.
/rest/v3/markets/{market_id}/orderbook endpoint now accepts a depth parameter with values up to 300./rest/v3/markets/{market_id}/orderbook endpoint is now 300.Check the documentation for more details.
POST /orders endpointGET /markets/sparkline/{window} endpointCheck the available documentation for this endpoint for more details.
best informationCheck the available documentation for this endpoint for more details.
X-FB-RECEIVE-WINDOW headerCheck the available documentation for more details.
order_type field
default_fees field
Both available in the response of the /rest/v3/markets endpoint
Check the available documentation for this endpoint for more details.
Added a new state to orders called PENDING_CANCEL for the following endpoints:
This state means that the order has been removed from the matching queue and is awaiting fund release to finalize the cancellation.
GET /otc/executions/{quote_id}STOP_LIMIT orderstime_in_force field, only available for LIMIT and STOP_LIMIT order typesrole field in the response of the /rest/v3/trades endpointCheck the available documentation for this endpoint for more details.
POST /v3/orders/batch