> ## Documentation Index > Fetch the complete documentation index at: https://docs.0xinsider.com/llms.txt > Use this file to discover all available pages before exploring further. # Stream > Receive live feed events through a Server-Sent Events connection and resume after reconnecting. Open one Server-Sent Events connection to receive new feed events, including large trades and wallet grade changes. Save each frame's ID after processing it so you can resume after a disconnect. For durable, paginated recovery of recorded large-trade events, use [Event replay](/api-reference/endpoint/get-event-replay-since). ## Parameters Filters are set when you connect and cannot be changed on an open connection. They apply to replayed frames and live frames alike, and a frame has to pass every filter you set. | Parameter | Description | | - | - | | `event` | A comma-separated list of frame types, such as `WhaleTradesInserted,wallet_grade_changed`. A frame passes only when its `type` is in your list. An unknown name returns `400` and names the value. `suspicious_trade_flagged` and `insider_radar_flag_raised` both select the flagged-trade frames; the frame itself still carries `type: insider_radar_flag_raised`. | | `condition_id` | One market, as the raw Polymarket `condition_id` or its `mkt_` id. A frame passes only when it carries a matching `condition_id`, so frames that carry none are excluded while this is set. | | `min_grade` | A wallet grade from `S` (best) to `F`. See [Grades](/concepts/grades). Grade-change and `LargeTradeInsertedV2` frames can pass. A large trade uses the trader grade captured when its event is published. Frames with no grade are excluded. | | `wallet` | A Polymarket wallet address. Only matching `LargeTradeInsertedV2` frames pass. | | `min_size` | A positive dollar amount. Only `LargeTradeInsertedV2` frames at least this large pass. | | `Last-Event-ID` | A request header holding the `seq` of the last frame you processed. The server replays everything after it, then goes live. | | `last_event_id`, `seq` | Query-string stand-ins for `Last-Event-ID`, for a client that cannot set the header. A value that is not a number attaches you live with no replay. | ## Frames | Frame | What it is | | - | - | | A `data:` frame | One JSON feed envelope, carrying `seq`, `published_at`, `type`, and the fields for that event type. | | The SSE `id:` | The frame's sequence number, shared across the whole cluster. It is the same number as `seq` in the JSON, and it stays valid across replicas and process restarts. | | `LargeTradeInsertedV2` | One trade, with a durable `trade.event_id` separate from the SSE sequence. The legacy count frame remains available. | | `LiveScoreChanged` | A complete score for `event_slug`, with optional API-Tennis `live_score.tennis_points`. The point group has its own revision and expiry. | | `event: resync` | A marker saying the server could not cover your resume point, or that its sequence numbering restarted. Refetch current state, keep the connection open, and store the marker's `id`. | | `event: error` | The last frame before the server closes the connection. See "When the stream ends on its own" below. | | `: keep-alive` | A comment line every 5 seconds on an idle connection, so nothing in between closes it. | A `resync` marker carries `completeness.status`: `truncated` when your resume point was older than the retained window, `lagged` when the stream fell behind, and `reset` when the server's sequence numbering restarted while your connection was open. Frames are never skipped without one. To receive tennis point updates, set `event=LiveScoreChanged` without market, wallet, grade, or size filters, which exclude these match-based frames. Apply the point group's own `source_revision`, `expires_at`, `set_number`, and `games`; Polymarket's enclosing score revision does not order point-only changes. Keep absent points or `serving_side` unknown, and resume missed frames with this stream's `Last-Event-ID` window. Compatible, unexpired `live_score.tennis_points` survives temporary supplemental snapshot unavailability or contention. Hide the group at expiry or when its player identity, set, game, or live state no longer matches. After a `reset` marker, the frames that follow carry lower `id` values than the frames before it. The marker's `id` is one below the first of the new numbering, so storing it as your cursor resumes with nothing missed; `from_sequence` is the last `id` you were sent and `to_sequence` is the first of the new numbering. ## How to reconnect without missing a frame 1. Open the stream with `curl -N` or a browser `EventSource`. Sending no resume cursor attaches you to live frames, with no replay and no error. 2. Store each frame's `id` after you have processed it. 3. On reconnect, send `Last-Event-ID: `. A browser `EventSource` does this for you. 4. Read the replayed frames in ascending `id`, then the live frames. There is no gap in `seq` between the two. 5. On `event: resync`, refetch current state, keep the connection open, and store the marker's `id` as your cursor. ## When the stream ends on its own The key is checked when you connect. Further checks become due every 30 seconds, with a 10-second credential cache TTL. Body polling and lookup I/O can delay terminal delivery. If the key is revoked, expired, or rotated, or the account that owns it is deleted, locked, or no longer subscribed, the stream ends with one final frame and the connection closes: ```text theme={null} event: error id: 9120 data: {"type":"error","error":{"code":"invalid_api_key","message":"The API key this stream was opened with is no longer valid (revoked, expired or rotated). The stream is closed; a reconnect with it is refused with 401 invalid_api_key.","doc_url":"https://docs.0xinsider.com/api-reference/introduction#authentication"},"retry":false} ``` `error` is the same object a reconnect is answered with in its response body, so one handler covers both cases. `retry: false` means a reconnect with the same credential is refused with that code: `invalid_api_key` (`401`), `subscription_required` (`402`), `forbidden` (`403`, a deleted account), `account_locked` (`423`), or `insufficient_scope` (`403`). Don't reconnect: fix the credential or the account first. There is one retryable case. If the credential store stops answering, protected delivery is permitted only within 90 seconds of the last confirmed check. Exhaustion produces code `rate_limit_unavailable`, reason `database_unavailable`, and `retry: true` when the body is polled; this does not guarantee wall-clock arrival of the frame. A healthy response paused beyond that period must complete one fresh check before protected delivery resumes. Reconnect after `error.retry_at` with `Last-Event-ID` set to that frame's `id`, and the replay covers what you missed. A browser `EventSource` also fires its own `error` event when the connection fails, with no `data`, so check `event.data` before you parse it. `@0xinsider/sdk` gives you the frame from `streamFeed()` as an envelope with `type` `error`, and `streamFeedResilient()` stops on the reconnect's `4xx`. ## Example ```bash theme={null} curl -N -H "Authorization: Bearer $OXINSIDER_API_KEY" \ -H "Last-Event-ID: $LAST_SEQ" \ "https://api.0xinsider.com/api/v1/stream?event=WhaleTradesInserted" ``` ## What it does not return * An open stream past the cap of 5 connections per key, or past the cluster-wide ceiling. Over either, you get `429` with `Retry-After`, so close the connections you are not reading. * An open stream while the admission or replay service is down. You get `503` with `Retry-After`: reconnect after the wait. * Another account's targeted events. Alert and following frames reach only the key that owns them. * Frames older than the retained window. You get one `resync` marker instead. * Recomputed Polymarket data. Each frame is the envelope 0xinsider broadcasts internally, forwarded as it is. * Anything at all on an idle connection, apart from the keep-alive comment every 5 seconds. * Further protected frames after revalidation refuses the credential or account. You get one `event: error` frame, and then the connection closes. Bytes already handed to HTTP cannot be recalled. ## OpenAPI ````yaml GET /api/v1/stream openapi: 3.1.0 info: x-generated-rate-limit-policy-from: web/src/lib/rate-limit-facts.ts via web/scripts/generate-api-policy.ts title: 0xinsider API description: >- Follow provider-exposed large-trade activity from Polymarket. Polymarket wallet-attributed trades can add grades, P&L, strategy, and diagnostic-score context when sufficient source data exists. Fields can be null or unavailable. Normal API requests use a 30-second server timeout that returns HTTP 408 Request Timeout with the standard error envelope (error.code request_timeout) when exceeded. Every /api/v1 failure answers that envelope, including a body that is not JSON or does not fit the request schema (400 invalid_body), a query or path value that does not parse (400 invalid_query, invalid_path), a missing Content-Type: application/json (415 unsupported_media_type), a body over 1048576 bytes (413 payload_too_large) and a method the path does not serve (405 method_not_allowed), each with error.param naming the field where one is known and meta.request_id equal to X-Request-Id. Unknown query names are ignored by default and reported in X-Query-Ignored, while X-Effective-Query lists the normalized names and values applied using form-urlencoded decoding, where + is a space; strict mode returns 400 bad_request with error.reason unknown_query_parameter before the handler runs, including for an unknown name with an incomplete percent escape. Every list operation clamps an out-of-range limit into its published minimum..maximum instead of refusing it (limit=0 reads one row, limit=500 reads the maximum), and X-Effective-Query reports the clamped value; only a limit that is not an integer is refused, with 400 invalid_query. Public REST /api/v1/* endpoints, excluding /api/v1/mcp, use Bearer-token based non-credentialed browser CORS: any Origin may call with Authorization, Content-Type, If-None-Match, Idempotency-Key, Mcp-Session-Id, Mcp-Protocol-Version, Last-Event-Id, and X-Query-Validation request headers. X-Query-Validation: strict opts into rejecting unknown query names; the default remains compatible. Remote MCP at /api/v1/mcp is non-credentialed, but still validates Origin against the 0xinsider/localhost allowlist per MCP Streamable HTTP DNS-rebinding guidance. Successful browser CORS preflight responses advertise Access-Control-Max-Age: 86400. Browser JavaScript may read RateLimit-Limit, RateLimit-Remaining, RateLimit-Reset, the legacy X-RateLimit-Limit, X-RateLimit-Remaining, X-RateLimit-Reset, Retry-After, ETag, X-Request-Id, X-Request-Cost, X-Usage-Accounting, X-Batch-RateLimit-Limit, X-Batch-RateLimit-Remaining, X-Batch-RateLimit-Reset, Mcp-Session-Id, X-Mcp-Error-Code, X-Query-Ignored, and X-Effective-Query response headers. Rate-limit headers describe the budget a request was counted against: the API key's per-minute window on an authenticated call, and the per-IP budget on a public route or on a refused credential (401, 402, 403, 423), so a client looping on a bad or lapsed key still sees how much room it has. Conditional GET: these operations return a weak ETag and answer If-None-Match with 304 Not Modified and an empty body: GET /api/v1/health, GET /api/v1/insider-radar, GET /api/v1/insider-radar/{id}, GET /api/v1/large-positions, GET /api/v1/leaderboard, GET /api/v1/leaderboard/trending, GET /api/v1/market/{condition_id}/candles, GET /api/v1/market/{condition_id}/flow, GET /api/v1/market/{condition_id}/intel, GET /api/v1/market/{condition_id}/snapshot, GET /api/v1/markets/explore, GET /api/v1/markets/sharp-money-flows, GET /api/v1/markets/smart-money-flows, GET /api/v1/pick-of-the-day, GET /api/v1/pick-of-the-day/archive, GET /api/v1/positions, GET /api/v1/sports-edge-observations, GET /api/v1/sports-edge-signals, GET /api/v1/trader/{address}, GET /api/v1/trader/{address}/context, GET /api/v1/trader/{address}/pnl, GET /api/v1/trader/{address}/position-timeline, GET /api/v1/traders/{trader}/position-timeline, GET /api/v1/large-trades, GET /api/v1/large-trades/history, GET /api/v1/large-trades/{id}, GET /api/v1/whale-trades, GET /api/v1/whale-trades/history, GET /api/v1/whale-trades/{id}, GET /api/v1/whale-trades/{id}/counterparties/executions, GET /api/v1/whale-trades/{id}/counterparties/executions/{execution_id}/makers. Credentialed first-party routes such as /api/keys, /api/billing, and auth endpoints remain restricted to configured 0xinsider origins. Protected V1 responses, except the zero-cost /api/v1/usage route, after handler execution carry X-Usage-Accounting: persisted, failed, or unknown. This reports the usage-record write; it does not change the handler result. Do not replay a successful mutation to repair an unknown usage record. Before execution, unavailable accounting capacity returns HTTP 503 with error.reason=request_accounting_unavailable; honor Retry-After. Public API responses add the browser-readable Server-Timing header: Processing time in milliseconds, for example api;dur=12.345. Includes API authentication, quota admission, handler work and response construction. Excludes network transit and streamed body or export-file transfer. The engineering budget is strictly below 250 ms; this header reports observations, not a latency guarantee or a new timeout. Requests through https://0xinsider.com/api/* append Server-Timing: web_api;dur= and X-Web-Request-Id for that web origin handler. The web clock includes upstream waiting and existing body construction, so do not add it to api;dur. It excludes platform routing, cold module initialization, CDN hits and later stream/file transfer. On a cache hit these headers describe the earlier origin fill, not the current request; static Markdown API routes have no fresh web clock. version: 1.0.0 contact: name: 0xinsider email: support@0xinsider.com url: https://0xinsider.com servers: - url: https://api.0xinsider.com description: >- Production (live data). Authenticate with a live key (oxi_sk_live_...); requires an active Pro subscription. A sandbox key (oxi_sk_test_...) is answered with 401 invalid_api_key and error.reason sandbox_api_key. - url: https://0xinsider.com/sandbox description: >- Sandbox. No credential required and no production data: every documented operation answers with its documented example or a deterministic sample of its response schema. GET /api/v1/stream is the one exclusion and answers 400 there, because a Server-Sent Events stream is a live connection rather than a body. Add ?sandbox_status= to receive one of the error responses the operation documents (for example 429 with Retry-After). Documented query parameters and JSON request bodies are checked against this document, the two context.md routes answer 200 text/markdown, GET /api/v1/trader/{address}/export/download answers its 302 with a Location the sandbox serves itself rather than an object store, and nothing is stored between requests. A sandbox key (oxi_sk_test_..., issued with no account by POST https://api.0xinsider.com/api/v1/agents/register) is optional: on an operation that requires a credential, a well-formed key is answered with X-Oxi-Sandbox-Key: valid and a malformed one with 401 invalid_api_key. security: - bearerAuth: [] - oauth2: - read tags: - name: Traders description: Traders, batch lookups, timelines, and export readiness. - name: Positions description: Current prediction-market position snapshots from backend-owned mirrors. - name: Large Positions description: Largest current open positions from graded traders (Polymarket-only). - name: Large trades description: Recent and historical large trades. - name: Leaderboard description: Ranked trader discovery and category/strategy leaderboards. - name: Pick of the Day description: >- One sourced sharp-money call a day: the side profitable wallets are backing, with pre-game odds, the holders, and the track record. - name: Games description: >- Sports and esports games: both sides, schedules, provider status and the Polymarket markets linked to each game. - name: Markets description: Market search, discovery, snapshots, and sharp-money flow. - name: Content description: Search across 0xinsider editorial content. - name: Suspicious trades description: Trades whose recorded suspicion score meets the live flag threshold. - name: Insider Radar description: >- Deprecated spelling of Suspicious trades; both operations stay live as aliases. - name: Events description: Durable public event replay streams. - name: Streaming description: Resumable real-time Server-Sent Events stream of live feed envelopes. - name: Webhooks description: Signed builder webhook destinations and delivery controls. - name: Usage description: Developer API budget and usage introspection. - name: Onboarding description: >- Self-serve agent registration: a sandbox key with no account, and the path to live access. - name: System description: Health and operational status checks. - name: MCP description: Remote Model Context Protocol transport. - name: Reports description: Daily, weekly, monthly, and trader export report snapshots. - name: Account description: Identify the account and credential authenticated for a paid API request. externalDocs: description: 0xinsider API docs url: https://docs.0xinsider.com paths: /api/v1/stream: get: tags: - Streaming summary: Resumable real-time event stream (SSE) description: >- Server-Sent Events stream of the live feed envelopes the platform already broadcasts (whale-trade pulses and other public/Pro feed events). Forwards the same backend-owned envelope shape as the internal feed; no provider data is recomputed. LiveScoreChanged frames may include optional live_score.tennis_points from API-Tennis in scoreboard order. This group has independent source_revision, observed_at, expires_at, set_number, and games; serving_side is omitted when unknown. Discard the group at expiry or set/game mismatch, and never compare its revision with the enclosing Polymarket score revision. Use event=LiveScoreChanged without market, wallet, grade, or size filters to receive these frames, which carry event_slug rather than condition_id. Authenticated via the oxi_sk Bearer key like every other /api/v1 endpoint, and limited to a small number of concurrent connections per API key and a cluster-wide ceiling across all keys (HTTP 429 with Retry-After when either cap is exceeded; HTTP 503 with Retry-After if a required Redis service is briefly unavailable). Each delivered frame carries an SSE id from one cluster-shared sequence that remains valid across backend replicas and process restarts. Reconnect with the Last-Event-ID header (or the last_event_id / seq query fallback) to replay the missed window before resuming live. When the requested resume point is older than the retained window, ahead of the current sequence, separated from live delivery by an uncovered gap, or the server's sequence counter restarts mid-stream (a completeness.status of truncated, lagged, or reset), the stream emits a resync marker event (event: resync) instead of silently skipping frames. Idle connections receive periodic ': keep-alive' comment lines. This is a long-lived response: keep the connection open and read frames as they arrive. Authorization is re-checked for as long as the stream is open: checks become due every 30 seconds and the credential cache has a 10-second TTL. Body polling and lookup I/O can delay terminal delivery. A key that is revoked, expired or rotated, or an account that is deleted, locked or no longer subscribed, ends the stream with one terminal 'event: error' frame and then closes the connection. The frame's JSON is { type: 'error', error: , retry: }; retry is false for every credential and account refusal (a reconnect is refused with the same 401, 402, 403 or 423), and true only when the credential store stopped answering (code rate_limit_unavailable with reason database_unavailable when protected-delivery grace expires 90 seconds after the last confirmed check; terminal arrival depends on body polling), in which case reconnect after the error's retry_at with Last-Event-ID set to the frame's id to resume. Replay uses the same ongoing authorization as live delivery. When a paused response resumes, due authorization and terminal controls run before another protected replay frame is emitted; bytes already handed to HTTP cannot be recalled. A healthy response paused beyond the grace period must complete one fresh check before protected delivery resumes. Compatible unexpired tennis_points survives temporary supplemental snapshot unavailability or contention; expiry and set/game mismatch still clear it. operationId: getStream parameters: - name: X-Query-Validation in: header required: false description: >- Opt into strict query-name validation. The default is compatible: unknown names are ignored and reported in X-Query-Ignored. With strict, an unknown name returns 400 bad_request with error.reason unknown_query_parameter before the handler runs, including when its percent escape is incomplete. schema: type: string enum: - strict - name: Last-Event-ID in: header required: false description: >- Cluster-shared sequence id of the last frame the client successfully processed. The stream replays the strictly-newer retained window before resuming live, even when reconnect routing changes replicas. A future or expired id emits resync. Browser EventSource clients send this automatically on reconnect. Omit to attach live from now. schema: type: string - name: last_event_id in: query required: false description: >- Query-string fallback for the Last-Event-ID resume cursor when the client cannot set the request header. Numeric cluster-shared sequence id; non-numeric or absent attaches live with no replay. A future or expired id emits resync. schema: type: string - name: seq in: query required: false description: >- Alias for last_event_id. Numeric cluster-shared sequence id to resume after; non-numeric or absent attaches live with no replay. A future or expired id emits resync. schema: type: string - name: event in: query required: false description: >- Optional per-connection subscribe-time filter. Comma-separated list of frame wire types (the SSE 'type' discriminator, e.g. WhaleTradesInserted,wallet_grade_changed); a frame passes only when its type is in the set. An unknown token returns HTTP 400 naming the offending value. An empty/all-whitespace value is treated as absent (no constraint). Applied to BOTH the replay window and live frames, and always AFTER the per-subscriber privacy gate, so it can never widen visibility beyond what the key may already see. Combines with condition_id and min_grade as a logical AND; an absent param adds no constraint. schema: type: string - name: condition_id in: query required: false description: >- Optional per-connection subscribe-time filter. Raw provider condition_id or mkt_-prefixed market id (normalized the same way the other v1 market endpoints normalize). A frame passes only when it carries a matching condition_id field; frames that carry no condition_id (e.g. whale-pulse events) are EXCLUDED while this is set. An empty value after normalization returns HTTP 400. Combines with event and min_grade as a logical AND; an absent param adds no constraint. schema: type: string - name: min_grade in: query required: false description: >- Optional per-connection subscribe-time filter. A frame passes only when it carries a grade field whose grade is better-or-equal to this minimum (S is best). Frames without a grade (every frame except wallet_grade_changed and LargeTradeInsertedV2) are excluded. An invalid grade returns HTTP 400. Combines with the other filters as a logical AND. schema: type: string enum: - S - A - B - C - D - F - name: wallet in: query required: false description: >- Filter LargeTradeInsertedV2 frames to this Polymarket wallet. Other event types have no wallet and are excluded when set. Applied to live and retained Redis frames after privacy checks. schema: type: string pattern: ^0x[a-fA-F0-9]{40}$ - name: min_size in: query required: false description: >- Filter LargeTradeInsertedV2 frames to trades at or above this USD notional. Positive decimal with at most six places; other event types are excluded when set. schema: type: string pattern: ^[0-9]+(\.[0-9]{1,6})?$ responses: '200': description: >- Open Server-Sent Events stream. The body is an unbounded text/event-stream of frames; each data frame is a JSON feed envelope (with cluster-shared seq, published_at, type, and event-specific fields) and carries an SSE id equal to seq. resync marker frames use event: resync for expired, future, lagged, sequence-reset, or otherwise uncovered cursors. A terminal event: error frame (id: , data: { type: 'error', error: { code, message, ... }, retry: }) ends the stream when the key or account it was opened with no longer authorizes it, or when the credential store has not confirmed it within 90 seconds; the connection closes after it. content: text/event-stream: schema: type: string description: >- Newline-delimited SSE frames: 'id: \ndata: \n\n' for events, 'event: resync\nid: \ndata: \n\n' for resync markers, 'event: error\nid: \ndata: \n\n' as the one terminal frame when the credential stops authorizing the stream, and ': keep-alive' comment lines while idle. examples: whaleEvent: summary: One live whale-trade event frame value: >+ id: 142 data: {"seq":142,"published_at":"2026-06-04T12:00:00Z","type":"WhaleTradesInserted","count":3} resyncMarker: summary: Resync marker after a too-old resume point value: >+ event: resync id: 9001 data: {"type":"resync","completeness":{"status":"truncated","reason":"The requested Last-Event-ID is outside the retained replay window; treat this as a full resync and refetch current state."},"from_sequence":12,"to_sequence":9001} credentialRevoked: summary: >- Terminal error frame after the key was revoked mid-stream; the connection closes after it value: >+ event: error id: 9120 data: {"type":"error","error":{"code":"invalid_api_key","message":"The API key this stream was opened with is no longer valid (revoked, expired or rotated). The stream is closed; a reconnect with it is refused with 401 invalid_api_key.","doc_url":"https://docs.0xinsider.com/api-reference/introduction#authentication"},"retry":false} headers: X-Query-Ignored: $ref: '#/components/headers/X-Query-Ignored' X-Effective-Query: $ref: '#/components/headers/X-Effective-Query' RateLimit-Limit: $ref: '#/components/headers/RateLimit-Limit' RateLimit-Remaining: $ref: '#/components/headers/RateLimit-Remaining' RateLimit-Reset: $ref: '#/components/headers/RateLimit-Reset' X-RateLimit-Limit: $ref: '#/components/headers/X-RateLimit-Limit' X-RateLimit-Remaining: $ref: '#/components/headers/X-RateLimit-Remaining' X-RateLimit-Reset: $ref: '#/components/headers/X-RateLimit-Reset' X-Request-Id: $ref: '#/components/headers/X-Request-Id' X-Usage-Accounting: $ref: '#/components/headers/X-Usage-Accounting' Server-Timing: $ref: '#/components/headers/Server-Timing' '400': description: Invalid request parameter content: application/json: schema: $ref: '#/components/schemas/ApiError' '401': description: Missing or invalid API key content: application/json: schema: $ref: '#/components/schemas/ApiError' '402': description: >- Active Pro subscription required. The key is valid but the account has no active Pro subscription; error.reason is subscription_inactive and error.message names the reactivation URL (https://0xinsider.com/billing). Permanent until a person reactivates: no Retry-After, never retry on a schedule. content: application/json: schema: $ref: '#/components/schemas/ApiError' '403': description: Account access denied content: application/json: schema: $ref: '#/components/schemas/ApiError' '423': description: Account is locked content: application/json: schema: $ref: '#/components/schemas/ApiError' '429': description: >- Rate limit exceeded. Three independent budgets. (1) 100 requests/minute per user (sliding window), on every authenticated route. (2) On the BATCH routes only: 2500 batch item units/minute per user, reserved before any item is executed. A batch with N requested items costs N item units, including duplicate and invalid items. 2500 = 100 requests x 25 items per batch, which is the most item work a key can buy through the request limiter at all: a caller may spend their entire 100-request minute on full 25-item batches without the item budget being what stops them. The REQUEST budget is the effective ceiling, and batching is never the more expensive choice. The item budget can still deny at a sliding-window boundary (both counters carry the previous window forward with a floor, and the item counter runs 25x the request counter), so honor a 429 from either. Over-quota batches return 429 with Retry-After plus RateLimit-Limit, RateLimit-Remaining, and RateLimit-Reset before any item work is done. (3) The monthly quota: Pro includes 500,000 and Max includes 2,000,000 authenticated requests per UTC calendar month. Over the plan's included requests: with pay as you go on, the excess bills at USD 0.20 per 1,000 on a monthly invoice, up to four times the included allowance (2,000,000 requests for Pro); without it, from October 1, 2026, the next request answers 429 rate_limited with error.reason monthly_quota_exceeded and a Retry-After to the month's reset. Pay-as-you-go accounts receive the same refusal at their ceiling. A refused request is not counted. Every authenticated response carries X-Monthly-Quota-Limit, X-Monthly-Quota-Remaining, and X-Monthly-Quota-Reset (unix seconds, the first of next month). (4) The per-address budget: 1200 requests/minute per IP, shared by every caller behind one address and counted before authentication, on every route. A 429 from it carries error.reason ip_rate_limited and describes that bucket in RateLimit-*; a throttled address (sustained over-limit traffic) carries error.reason ip_throttled with a Retry-After of minutes to days, and a request before it does not shorten the cooldown. Every 429 is the standard error envelope with meta.request_id equal to X-Request-Id. headers: Retry-After: description: Seconds until rate limit resets. schema: type: integer RateLimit-Limit: $ref: '#/components/headers/RateLimit-Limit' RateLimit-Remaining: $ref: '#/components/headers/RateLimit-Remaining' RateLimit-Reset: $ref: '#/components/headers/RateLimit-Reset' X-RateLimit-Limit: schema: type: integer X-RateLimit-Remaining: schema: type: integer X-RateLimit-Reset: schema: type: integer X-Request-Id: schema: type: string content: application/json: schema: $ref: '#/components/schemas/ApiError' '500': description: Unexpected server error content: application/json: schema: $ref: '#/components/schemas/ApiError' '503': description: >- A required Redis-backed admission or shared replay service is unavailable; retry after the outage cooldown headers: Retry-After: description: >- Seconds until the middleware will probe the Redis-backed rate limiter again. schema: type: integer X-Request-Id: schema: type: string content: application/json: schema: $ref: '#/components/schemas/ApiError' x-codeSamples: - lang: curl label: cURL source: |- curl -N -sS \ -H "Authorization: Bearer $OXINSIDER_API_KEY" \ 'https://api.0xinsider.com/api/v1/stream' - lang: curl label: cURL (resume) source: |- curl -N -sS \ -H "Authorization: Bearer $OXINSIDER_API_KEY" \ -H 'Last-Event-ID: 100' \ 'https://api.0xinsider.com/api/v1/stream' components: headers: X-Query-Ignored: description: >- Comma-separated, percent-encoded query names the operation did not publish and therefore ignored in compatible mode. Names are sorted and de-duplicated. schema: type: string X-Effective-Query: description: >- Normalized, percent-encoded query string containing only the recognized names and values applied by the operation. Query names and values use form-urlencoded decoding, where + is a space. Repeated names are retained and sorted by name. A limit outside the operation's published minimum..maximum is reported as the clamped value the page used, so the header states the page size served rather than the one requested. schema: type: string RateLimit-Limit: description: >- Request limit of the budget this request was counted against, for its current window: the API key's per-minute sliding window on an authenticated call; the per-IP budget on a public route and on a refused credential (401, 402, 403, 423), which never reaches the per-key limiter. Standard RateLimit header spelling. schema: type: integer example: 100 RateLimit-Remaining: description: >- Requests remaining in that budget's current window after this response. Standard RateLimit header spelling. schema: type: integer example: 84 RateLimit-Reset: description: >- Seconds until that budget's current window resets. Standard RateLimit header spelling. schema: type: integer example: 42 X-RateLimit-Limit: description: >- Request limit of the budget this request was counted against, for its current window: the API key's per-minute sliding window on an authenticated call; the per-IP budget on a public route and on a refused credential (401, 402, 403, 423). schema: type: integer example: 100 X-RateLimit-Remaining: description: Requests remaining in that budget's current window after this response. schema: type: integer example: 84 X-RateLimit-Reset: description: Unix timestamp when that budget's current window resets. schema: type: integer example: 1710772860 X-Request-Id: description: >- Server-generated request identifier for support and tracing. On every /api/v1 response, including 304, 408, CORS preflights and every error, and always equal to meta.request_id in the body. It is the key of the request's usage accounting row and of every log line the request emitted, so quote either form to support. A client-supplied X-Request-Id request header is ignored: the value is never adopted or echoed. schema: type: string example: req_550e8400 X-Usage-Accounting: description: >- Usage-record persistence for a protected V1 handler response. persisted: confirmed row; failed: write failed; unknown: completion could not be confirmed. Independent of handler success; do not replay successful mutations to repair accounting. Absent before accounting admission and on public routes. The zero-cost /api/v1/usage route also omits it. schema: type: string enum: - persisted - failed - unknown Server-Timing: description: >- Processing time in milliseconds, for example api;dur=12.345. Includes API authentication, quota admission, handler work and response construction. Excludes network transit and streamed body or export-file transfer. The engineering budget is strictly below 250 ms; this header reports observations, not a latency guarantee or a new timeout. schema: type: string example: api;dur=12.345 schemas: ApiError: type: object required: - object - error - meta properties: object: type: string const: error error: type: object required: - code - message properties: code: type: string description: >- FROZEN: an existing value never changes meaning. request_timeout (408, #16146) was added the way insufficient_scope was: the handler did not answer inside the server's 30-second timeout. Retry-After and retry_at ride on it only for a safe method (GET, HEAD); a timed-out mutation may have completed, so check its state and reuse its Idempotency-Key. enum: - bad_request - invalid_api_key - subscription_required - forbidden - insufficient_scope - not_found - account_locked - rate_limited - rate_limit_unavailable - internal_error - request_timeout message: type: string doc_url: type: string param: type: string retry_at: type: string format: date-time description: >- The recommended next request instant (RFC3339), always in the future. Present on every retryable error: `pick_not_released`, `rate_limited`, `rate_limit_unavailable`, and `read_model_warming`. Omitted otherwise. The absolute twin of `Retry-After`; prefer the header for the sleep duration. For `pick_not_released`, the earliest of the next scheduled release, the next automatic selector attempt, the operating-window start, or about 60 seconds. See that response. freshness: $ref: '#/components/schemas/FreshnessFailure' reason: type: string enum: - cursor_expired - unknown_endpoint - pick_not_released - trader_not_tracked - read_model_warming - database_unavailable - request_accounting_unavailable - idempotency_in_progress - webhook_delivery_in_progress - webhook_secret_rotation_not_prepared - webhook_secret_rotation_overlap_active - sandbox_api_key - api_key_in_query - subscription_inactive - monthly_quota_exceeded - invalid_query - unknown_query_parameter - invalid_path - invalid_body - unsupported_media_type - payload_too_large - method_not_allowed - ip_rate_limited - ip_throttled - export_expired - freshness_ceiling_unsatisfied description: >- ADDITIVE (#7209). The specific, actionable cause behind `code`, when there is one more specific than the code itself. `code` keeps its published values, so existing clients are unaffected; new clients branch on `reason`. Omitted when the code already says everything we know. pick_not_released: no Pick of the Day is published for the current product day; schedule one request against retry_at instead of polling. unknown_endpoint: the PATH is not a route on this API -- read GET /api/v1, do not retry. trader_not_tracked: the wallet is real and the URL is right, but the trader is outside the HOT/WARM sync tiers -- stop asking for this wallet. cursor_expired: pagination went stale mid-walk -- re-request the first page and continue. read_model_warming: the requested endpoint cannot serve its read model yet; exact causes are endpoint-specific and can include a cold or contended refresh or a dependency that prevented refresh. database_unavailable: the API's database or its connection pool is temporarily unreachable (a connection-class failure, not a query fault); code stays rate_limit_unavailable, nothing is rate-limited, retry after Retry-After / retry_at. idempotency_in_progress: retain the exact Idempotency-Key and request body, then retry shortly. webhook_delivery_in_progress: retry the URL or signing-secret configuration change after the destination's active request completes. request_accounting_unavailable: accounting capacity is unavailable before the handler executes; retry after Retry-After / retry_at. sandbox_api_key: the credential is a sandbox key (oxi_sk_test_) from POST /api/v1/agents/register, which only the sandbox server accepts -- call the sandbox base URL with it, or get a live key or OAuth access token; do not retry it here. api_key_in_query: the key was sent as a ?token= query parameter, which no route reads because URLs land in logs and history; the key itself was not checked -- resend it as Authorization: Bearer. subscription_inactive: the key is valid but the account's Pro subscription has lapsed (402 subscription_required); permanent until a person reactivates at https://0xinsider.com/billing, which the message names -- stop retrying on a schedule and surface the link. The key owner is emailed once per lapse. monthly_quota_exceeded: the account has used the requests Pro includes for the UTC calendar month (429 rate_limited); retry_at and Retry-After name the first of next month, the only retry that can succeed, and the message names https://0xinsider.com/developers, where pay as you go for requests over the quota is turned on. The X-Monthly-Quota-Limit, X-Monthly-Quota-Remaining and X-Monthly-Quota-Reset headers on every authenticated response say how close the account is. invalid_query, invalid_path, invalid_body (400 bad_request, #16146): a query parameter, a path segment or the JSON body did not parse or does not fit the route's schema, so no handler ran; param names the field when the parser named one (a query key, a path segment, a JSON path such as traders[0], or body); fix the request, never retry it as sent. unsupported_media_type (415 bad_request, param content-type): send the body with Content-Type: application/json. payload_too_large (413 bad_request, param body): the body is over 1048576 bytes. method_not_allowed (405 bad_request): the path is a route but not with this method; the Allow header names the methods it serves. ip_rate_limited (429 rate_limited, #16380): the per-address budget every caller behind one IP shares, counted before authentication, is spent; not the key's own window, and the RateLimit-* headers describe that bucket. ip_throttled (429 rate_limited): the address is in a cooldown after sustained over-limit traffic; Retry-After is minutes to days, and a request before it does not shorten the cooldown. meta: $ref: '#/components/schemas/ResponseMeta' FreshnessFailure: type: object required: - max_age_s - data_quality_status properties: max_age_s: type: integer format: int64 minimum: 0 description: The caller's requested whole-response freshness ceiling in seconds. actual_age_s: type: integer format: int64 minimum: 0 description: >- Age in seconds of the oldest stored data_quality.as_of clock, when one is available. as_of: type: string format: date-time description: >- The oldest stored data-quality clock used to calculate actual_age_s, when one is available. data_quality_status: type: string enum: - fresh - partial - unknown - untracked - unavailable description: >- The trader body's whole-response data-quality status. Only fresh can satisfy max_age_s. ResponseMeta: type: object required: - request_id - cached - cost properties: request_id: type: string description: >- Unique request ID (req_ prefix). The same value as the X-Request-Id response header, the request's usage accounting row and its log lines. cached: type: boolean cache_age_s: type: integer description: >- Cache age in seconds. Omitted when the response was not cached, and also when it was cached but its age cannot be established (an entry stored before its cache carried a computed instant). Never a placeholder: an unknown age is reported as no value rather than as the cache TTL. cost: type: integer description: >- Advisory request weight (relative compute cost). 1 for simple reads; higher for heavier endpoints. Not a credit/price. ranking_generation: type: integer description: >- Committed PostgreSQL-owned leaderboard generation for the returned rows and cursor. Present on GET /api/v1/leaderboard; omitted on endpoints that do not read this ranking. ranking_as_of: type: string format: date-time description: >- Authoritative RFC3339 timestamp from cache_generations.updated_at for ranking_generation. It is read in the same repeatable-read snapshot as the leaderboard rows and is not request time, cache write time, or row insertion order. directional_source: type: string enum: - live - degraded description: >- Which path produced the team-directional read on this response. Only present on endpoints that compute one (today: GET /api/v1/sports-edge-signals). "live" means the read RAN. "degraded" means it FAILED, so nothing was measured and the ranking fell back to raw conviction. The flag describes the READ, not its consequence: a read that ran and found nothing groupable also leaves the directional fields null, and that is honestly "live" -- the per-signal nulls already say "nothing to enrich here", so this snapshot-level flag carries only what they cannot, namely whether the read ran at all. A degraded response is cached on the shorter degraded TTL so it self-heals. Reported SEPARATELY from ranking_source because the two degradations are independent -- a sharp-money DB miss weakens the ranking DATA, a directional failure removes a ranking WEIGHT -- and a consumer down-weighting a degraded response needs to know which input it lost. Omitted on endpoints that compute no directional read. ranking_source: type: string enum: - live - db_only description: >- Which ranking-data path produced this response. Only present on endpoints that can degrade a ranking (today: GET /api/v1/sports-edge-signals). "live" is the normal path (the current holder pile from the provider batch); "db_only" is the degraded fallback (a truthful but weaker trader_markets ranking) served when the live sharp-money ranking batch is unavailable (a sharp-money DB read failure, not a Polymarket outage) and cached on a shorter TTL, so a consumer can down-weight or skip it. Omitted on endpoints that never degrade. category_skill_source: type: string enum: - live - partial - degraded - unavailable description: >- Whole filtered snapshot category-evidence status before pagination. Operational live always remains partial source coverage. category_skill_model_version: type: string category_skill_taxonomy_version: type: string category_skill_platform: type: string const: polymarket category_skill_scope: type: string const: observed_goldsky_primary_taker_fill category_skill_source_coverage: type: string enum: - partial_whale_threshold_fills - graded_wallet_fills category_skill_observation_started_at: type: string format: date-time category_skill_model_operationally_degraded: type: boolean description: >- Whole-model operational readiness captured with the category model snapshot. Present on category-enriched responses even when the filtered signal list is empty. When true, category_skill_source is degraded and sports-edge-signals uses the shorter degraded cache TTL. category_skill_status_counts: type: object required: - live - insufficient - stale - unknown - degraded properties: live: type: integer minimum: 0 insufficient: type: integer minimum: 0 stale: type: integer minimum: 0 unknown: type: integer minimum: 0 degraded: type: integer minimum: 0 category_skill_base_payload_hash: type: string pattern: ^[0-9a-f]{64}$ description: >- SHA-256 of the funded signal membership/order/rank/cursor vector immediately before category-skill enrichment. Sports-edge-signals only. category_skill_enriched_base_payload_hash: type: string pattern: ^[0-9a-f]{64}$ description: >- Independent SHA-256 recomputation over the same base fields immediately after category-skill enrichment. Equality with category_skill_base_payload_hash proves shadow enrichment did not change funded inputs. Sports-edge-signals only. securitySchemes: bearerAuth: type: http scheme: bearer description: >- Legacy default or named integration API key, or OAuth 2.1 access token, in the Authorization header as `Bearer oxi_sk_live_...` or `Bearer oxi_at_...`. Default keys retain full access; integration keys are limited to their approved read, webhooks, export and usage scopes and expire within 90 days. All credentials share the owner's account limits. Data calls require an active Pro subscription and return live data. A 401 carries WWW-Authenticate: Bearer resource_metadata="https://api.0xinsider.com/.well-known/oauth-protected-resource" (RFC 6750 section 3, RFC 9728). oauth2: type: oauth2 description: >- OAuth 2.1 authorization code flow with PKCE S256 for apps and MCP clients acting for a user. Public clients only (no client secret): register with RFC 7591 at https://api.0xinsider.com/oauth/register or present an https client ID metadata document URL as client_id. Authorization server metadata: https://api.0xinsider.com/.well-known/oauth-authorization-server. The access token (oxi_at_..., one hour) is sent as `Authorization: Bearer`; refresh tokens rotate on every use. A route outside the token's scopes answers 403 insufficient_scope. Walkthrough: https://0xinsider.com/auth.md. flows: authorizationCode: authorizationUrl: https://0xinsider.com/oauth/authorize tokenUrl: https://api.0xinsider.com/oauth/token refreshUrl: https://api.0xinsider.com/oauth/token scopes: read: >- Read markets, traders, large trades, positions, reports, search, the event stream and every MCP tool webhooks: Create, list, verify, rotate and delete webhook endpoints export: Start, poll and download trader exports usage: Read the caller's API usage counters ```` This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.