> ## Documentation Index > Fetch the complete documentation index at: https://docs.kairos.trade/llms.txt > Use this file to discover all available pages before exploring further. # Get up to 200 candle series in one request > Batched form of `GET /v1/candles`: fetches up to 200 independent `(provider, contract_id, timeframe_seconds, start, end, outcome)` series concurrently and returns them in request order, with **per-index partial failure** — one item's validation or fetch error does not fail the others. Each item is validated and aligned exactly as in `GET /v1/candles` (same clamping, rollup, and inversion rules). Body accepts either a `requests` array or a legacy `items` alias; if `requests` is non-empty it is used, otherwise `items` is used. `timeframe_seconds` and `outcome` in each item may be a JSON number or a numeric string. The request body is capped at 4 MiB. **Rate limiting / cost.** Admitted at the base cost of 1 `light`-bucket unit. Larger batches — more items, wider windows, finer timeframes — consume more of your quota; any excess is charged as a deferred extra charge against your next request. **Caching.** Batch responses are always `Cache-Control: no-store` and never carry an ETag. **Binary format.** Pass `?fmt=binary` or `Accept: application/x-kairos-candles` to receive a single multi-result binary frame (`num_results = len(requests)`) — one columnar result section per request index, in input order. An item that failed sets the high bit (`0x80000000`) of its section's `count`, so it stays distinguishable from a genuinely empty window; mask the bit off to read the (always zero) count. Per-item error messages are only available in the JSON response. **Failure semantics.** Returns `500` only when *every* item in the batch failed. Otherwise `200` with a mix of populated and empty/errored entries. ## OpenAPI ````yaml /openapi/market-data-api.yaml post /v1/candles/batch openapi: 3.1.0 info: title: Kairos Market Data API version: 1.2.0 summary: Read-only prediction and perpetual market data across venues. description: > The Kairos Market Data API provides read-only access to normalized prediction-market data across every venue Kairos integrates: OHLCV candles, the trade tape, market metadata, resolution lifecycles, and latest mark prices. ## No signup required Every endpoint works **without credentials** on the anonymous free tier, rate-limited per source IP (120 light units/min, 20 heavy units/min). Try any request from the interactive docs at [app.kairos.trade/docs/market-data](https://app.kairos.trade/docs/market-data) or straight from curl. For production budgets (1,200 light and 150 heavy units/min by default, raisable per credential), request an API key from the Kairos team. ## Authentication Three modes, resolved per request: - **Anonymous** — no credential headers at all. Free tier, IP-keyed limits. - **API key** — `X-Client-Id` + `X-Api-Key` + `X-Api-Secret` headers (all three; partial header sets are rejected with 401 rather than degrading to anonymous). - **First-party session** — `Authorization: Bearer` JWT (Kairos web app sessions only; not offered to API consumers). Resolution order is fixed: an `Authorization` header takes the bearer path, any of the three API-key headers takes the API-key path, and only a request carrying neither is eligible for the anonymous tier. On the API-key path, an unreachable credential store surfaces on **any** endpoint as `500` with `code: internal` and the message `authentication backend unavailable`. A source IP outside a credential's whitelist is `403 ip_not_whitelisted`; that is the only 403 this service emits. ## Rate limiting Weighted sliding-window budgets over one minute, in **units**, across two independent buckets: light (candles, metadata, resolutions, trade metrics) and heavy (trade history, marks, perpetual snapshots). Most requests cost 1 unit; larger requests may consume more of your quota — each operation's `x-kairos-rate-limit` states its exact cost model. Every admitted response carries `X-RateLimit-Bucket`, `X-RateLimit-Tier`, `X-RateLimit-Remaining`, and `X-RateLimit-Reset`; 429s add `Retry-After`. A separate per-IP admission gate runs *before* authentication to keep a flood off shared infrastructure. It also answers 429 `rate_limited`, but with `Retry-After: 60` and no `X-RateLimit-*` headers. Redis backs the weighted limiter and it fails **closed**: if it is unreachable the request is rejected with 503 `rate_limiter_unavailable` rather than let through unmetered. ## Conventions - **Prediction-market prices** are on the 0–100 scale (implied probability × 100) in the prediction candle, trade, and mark endpoints. Prediction USD notional = `size × price / 100`. - **Perpetual prices are direct venue prices**, not probabilities and not 0–100 values. Perpetual quantities retain their declared native unit (`base_asset` or `contracts`). Never divide a perpetual price by 100 or compute contract notional without the instrument's authoritative contract multiplier; the snapshot does not synthesize a cross-venue notional. - **Hyperliquid is two products on one venue.** Its HIP-4 outcome markets are prediction markets, read with `provider=hyperliquid` on the prediction endpoints; its perpetuals are bare coins read under `/v1/perpetuals/hyperliquid/{instrument}/snapshot`, where the instrument is the canonical id `hl-mainnet--` (e.g. `hl-mainnet-btc-usdt`). Different id schemes, different price semantics; they are never interchangeable. - **Errors** all use one envelope: `{"error": {"code": "...", "message": "..."}}`. - **Providers** are case-insensitive; `kalshi_offchain` aliases `kalshi`, `dome` aliases `polymarket`, and `opinion` is resolvable for historic reads only. contact: name: Kairos url: https://app.kairos.trade/docs/market-data termsOfService: https://kairos.trade/terms servers: - url: https://md.kairos.trade description: Production - url: https://staging-md.kairos.trade description: Staging security: - {} - apiKeyClientId: [] apiKeyKey: [] apiKeySecret: [] tags: - name: Candles description: OHLCV candle series (1s → 1d), JSON or compact columnar binary. - name: Trades description: Trade tape and aggregate volume metrics. - name: Markets description: Market metadata, batch lookup, identifier resolution, and enumeration. - name: Resolutions description: Resolution outcomes and UMA-style lifecycle state/timelines. - name: Marks description: Latest mark (last trade price) per outcome token. - name: Perpetuals description: Live perpetual books, trades, candles, funding, and market state. Beta. - name: Status description: Health and readiness probes. paths: /v1/candles/batch: post: tags: - Candles summary: Get up to 200 candle series in one request description: | Batched form of `GET /v1/candles`: fetches up to 200 independent `(provider, contract_id, timeframe_seconds, start, end, outcome)` series concurrently and returns them in request order, with **per-index partial failure** — one item's validation or fetch error does not fail the others. Each item is validated and aligned exactly as in `GET /v1/candles` (same clamping, rollup, and inversion rules). Body accepts either a `requests` array or a legacy `items` alias; if `requests` is non-empty it is used, otherwise `items` is used. `timeframe_seconds` and `outcome` in each item may be a JSON number or a numeric string. The request body is capped at 4 MiB. **Rate limiting / cost.** Admitted at the base cost of 1 `light`-bucket unit. Larger batches — more items, wider windows, finer timeframes — consume more of your quota; any excess is charged as a deferred extra charge against your next request. **Caching.** Batch responses are always `Cache-Control: no-store` and never carry an ETag. **Binary format.** Pass `?fmt=binary` or `Accept: application/x-kairos-candles` to receive a single multi-result binary frame (`num_results = len(requests)`) — one columnar result section per request index, in input order. An item that failed sets the high bit (`0x80000000`) of its section's `count`, so it stays distinguishable from a genuinely empty window; mask the bit off to read the (always zero) count. Per-item error messages are only available in the JSON response. **Failure semantics.** Returns `500` only when *every* item in the batch failed. Otherwise `200` with a mix of populated and empty/errored entries. operationId: postCandlesBatch requestBody: required: true content: application/json: schema: $ref: '#/components/schemas/CandleBatchRequest' example: requests: - provider: polymarket contract_id: >- 21742633143463906290569050155826241533067272736897614950488156847949938836455 timeframe_seconds: 3600 start: '2026-07-15T00:00:00Z' end: '2026-07-16T00:00:00Z' outcome: 0 - provider: kalshi contract_id: KXHIGHNY-26JUL15-T50 timeframe_seconds: 60 start: '2026-07-15T12:00:00Z' end: '2026-07-15T13:00:00Z' responses: '200': description: >- Per-index results (populated, empty, or errored). Order matches the request array. content: application/json: schema: $ref: '#/components/schemas/CandleBatchResponse' examples: mixedResult: value: results: - index: 0 candles: - contract_id: >- 21742633143463906290569050155826241533067272736897614950488156847949938836455 timeframe_seconds: 3600 bucket_start: '2026-07-15T00:00:00+00:00' open: 61.5 high: 63 low: 60.8 close: 62.4 volume: 184230 - index: 1 candles: [] error: unknown provider "acme" application/x-kairos-candles: schema: $ref: '#/components/schemas/CandleBinaryFrame' '400': description: Malformed body, empty/oversized `requests` array. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' examples: tooLarge: value: error: code: invalid_request message: batch size 250 exceeds maximum 200 empty: value: error: code: invalid_request message: requests array is required '401': $ref: '#/components/responses/Unauthorized' '403': $ref: '#/components/responses/IPNotWhitelisted' '429': $ref: '#/components/responses/RateLimited' '500': description: Every item in the batch failed (validation or fetch). content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' examples: default: value: error: code: internal message: all batch items failed '503': $ref: '#/components/responses/RateLimiterUnavailable' components: schemas: CandleBatchRequest: type: object description: > Request body of `POST /v1/candles/batch`. Provide either `requests` or the legacy `items` alias — if `requests` is non-empty it takes precedence, otherwise `items` is used. At least one entry is required across the two fields, and the effective array may not exceed 200 entries. properties: requests: type: array items: $ref: '#/components/schemas/CandleBatchItem' maxItems: 200 items: type: array items: $ref: '#/components/schemas/CandleBatchItem' maxItems: 200 description: >- Legacy alias for `requests`, used only when `requests` is absent or empty. CandleBatchResponse: type: object description: Response body of `POST /v1/candles/batch` in JSON mode. required: - results properties: results: type: array items: $ref: '#/components/schemas/CandleBatchResult' CandleBinaryFrame: type: string format: binary description: > Columnar binary candle frame (`Content-Type: application/x-kairos-candles`), returned by both candle endpoints when negotiated via `?fmt=binary` or an `Accept: application/x-kairos-candles` header. Byte layout (little-endian): `u8 magic=0xCA, u8 version=1, u16 num_results`, then for each section `u32 index, u32 count`, followed by six columnar arrays of length `count`: `u32[] t` (bucket start, epoch seconds), `u16[] o,h,l,c` (prices scaled ×100 of the 0-100 float, clamped to [0, 10000]), and `i64[] vol` (volume ×100). In a batch frame, a failed item sets the high bit (`0x80000000`) of its section's `count` — mask it off before using the count, and treat a set bit as "this series errored" rather than "no data". The error message itself is only available in the JSON response. ErrorResponse: type: object description: Canonical error envelope emitted by every Market Data API endpoint. required: - error properties: error: type: object required: - code - message properties: code: type: string description: Machine-readable error code. example: invalid_request message: type: string description: Human-readable detail. example: market_ids exceeds maximum 200 CandleBatchItem: type: object description: >- One series request inside a batch. Same validation rules as the `GET /v1/candles` query parameters. required: - provider - contract_id - timeframe_seconds - start - end properties: provider: type: string enum: - kalshi - polymarket - opinion - predictfun - hyperliquid - dome - kalshi_offchain example: polymarket contract_id: type: string maxLength: 128 example: >- 21742633143463906290569050155826241533067272736897614950488156847949938836455 timeframe_seconds: description: >- Bucket width in seconds. Accepted as a JSON number or a numeric string. Valid values 1, 60, 300, 900, 3600, 14400, 86400. oneOf: - type: integer - type: string example: 3600 start: type: string description: >- Same accepted formats as `GET /v1/candles`'s `start` parameter (RFC 3339, bare datetime, or bare date). example: '2026-07-15T00:00:00Z' end: type: string description: Same accepted formats as `GET /v1/candles`'s `end` parameter. example: '2026-07-16T00:00:00Z' outcome: description: >- Zero-based outcome index. Accepted as a JSON number or a numeric string. Defaults to 0. oneOf: - type: integer - type: string default: 0 example: 0 CandleBatchResult: type: object description: Result for one input item, at the same array position as the request. required: - index - candles properties: index: type: integer description: Zero-based position in the input `requests`/`items` array. example: 0 candles: type: array items: $ref: '#/components/schemas/CandleObject' description: >- Empty when the item's window was authoritatively empty OR when the item errored (check `error`). error: type: string description: >- Present only when this item failed validation or fetch. When present, `candles` is always `[]`. example: unknown provider "acme" CandleObject: type: object description: One OHLCV bucket. Prices are on a 0-100 scale (percentage probability). required: - contract_id - timeframe_seconds - bucket_start - open - high - low - close - volume properties: contract_id: type: string description: >- Echo of the request's `contract_id` (the caller-supplied id, not the resolved canonical market id). example: >- 21742633143463906290569050155826241533067272736897614950488156847949938836455 timeframe_seconds: type: integer enum: - 1 - 60 - 300 - 900 - 3600 - 14400 - 86400 description: Bucket width in seconds, echoing the request. example: 3600 bucket_start: type: string description: >- Bucket start timestamp, UTC, always rendered with an explicit `+00:00` offset (not `Z`). example: '2026-07-15T00:00:00+00:00' open: type: number format: double minimum: 0 maximum: 100 description: Opening price (0-100 scale). example: 61.5 high: type: number format: double minimum: 0 maximum: 100 description: High price in the bucket (0-100 scale). example: 63 low: type: number format: double minimum: 0 maximum: 100 description: Low price in the bucket (0-100 scale). example: 60.8 close: type: number format: double minimum: 0 maximum: 100 description: Closing price (0-100 scale). example: 62.4 volume: type: integer format: int64 description: Traded volume in the bucket, provider-native units. example: 184230 token_id: type: string description: >- Resolved per-outcome CLOB token id. Omitted entirely from the JSON object when empty. example: '704721957297303853272349184' responses: Unauthorized: description: > No valid credential presented. Every path returns `code: unauthorized`; the triggers differ: - An `Authorization` header on a deployment where bearer auth is not enabled (`bearer authentication is not enabled`). - A bearer token that fails verification (`invalid or expired session`), or whose token/session has been revoked (`session has been revoked`). - An incomplete or wrong API-key header set. Sending *any* of `X-Client-Id`, `X-Api-Key`, `X-Api-Secret` commits the request to the API-key path, so a partial set is rejected rather than silently degrading to anonymous limits (`valid X-Client-Id, X-Api-Key, and X-Api-Secret headers are required`). - No credential material at all while the anonymous free tier is unavailable — switched off for the deployment, forced off through the runtime kill switch, or not yet resolvable on a replica that has never read the flag (it fails closed). headers: WWW-Authenticate: schema: type: string description: >- `Bearer realm="market-data-api"` on bearer-verification failure, `APIKey realm="market-data-api", header="X-Client-Id, X-Api-Key, X-Api-Secret"` on missing/invalid API-key credentials. Absent on the revoked-session and bearer-disabled paths. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: unauthorized message: >- valid X-Client-Id, X-Api-Key, and X-Api-Secret headers are required IPNotWhitelisted: description: API key presented from a source IP not in that credential's whitelist. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: ip_not_whitelisted message: source IP not in credential whitelist RateLimited: description: > Two different gates return this status, both with `code: rate_limited`: - **Per-credential (or per-IP, when anonymous) weighted budget exhausted.** The message is `rate limit exceeded`, or, for anonymous callers, `anonymous rate limit exceeded — request an API key for higher limits (…)`. `Retry-After` is the whole number of seconds until the current one-minute window rolls over, floored at 1. All four `X-RateLimit-*` headers are present. - **Per-IP admission gate.** An in-process, pre-authentication guard that keeps a flood off the shared credential cache and database. The message is `too many requests from this address` and `Retry-After` is a flat `60`. Because this fires *before* authentication, the response carries **no** `X-RateLimit-*` headers at all. headers: Retry-After: schema: type: integer description: >- Seconds until the caller may retry — window-rollover seconds for the weighted limiter, a flat 60 for the IP admission gate. X-RateLimit-Bucket: schema: type: string enum: - light - heavy X-RateLimit-Tier: schema: type: string enum: - anonymous - api-key - user X-RateLimit-Remaining: schema: type: integer X-RateLimit-Reset: schema: type: integer description: Unix timestamp (seconds) when the current window resets. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: rate_limited message: rate limit exceeded RateLimiterUnavailable: description: > The rate limiter is unreachable. Fails CLOSED — the request is rejected rather than let through unmetered. content: application/json: schema: $ref: '#/components/schemas/ErrorResponse' example: error: code: rate_limiter_unavailable message: rate limiter temporarily unavailable securitySchemes: apiKeyClientId: type: apiKey in: header name: X-Client-Id description: >- Credential client id (`kairos_ck_...`). Must be sent together with X-Api-Key and X-Api-Secret. apiKeyKey: type: apiKey in: header name: X-Api-Key description: >- 64-char hex API key. Must be sent together with X-Client-Id and X-Api-Secret. apiKeySecret: type: apiKey in: header name: X-Api-Secret description: >- 64-char hex API secret. Must be sent together with X-Client-Id and X-Api-Key. ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.