Rate Limits

Sliding-window, per-key, per-endpoint-category. Independent bucket accounting — read quota exhaustion never blocks trading, and vice versa. Every response returns X-RateLimit-* headers so client-side throttling can be exact rather than heuristic.

Read-live baseline & tier multipliers

Approved builder keys start at self_serve and auto-graduate to read_live at $100 30-day attributed volume; higher tiers multiply the per-category budgets. Sandbox keys (tier=self_serve) share the same host and quotas — the $10 per-order + $100 daily notional caps handle safety, not rate throttling.

Endpoint category	Read-live 1×	Genesis 10×	Partner up to 20× / negotiable
Read (GET /v1/markets, GET /v1/data/history, orderbook, candles)	fair-use	10× fair-use	uncapped for committed contracts
Account (GET /v1/builders/me, attribution, revenue)	fair-use	10× fair-use	uncapped for committed contracts
Trade (order placement + cancel, single + batch)	fair-use	10× fair-use	MM-tier burst, contract-defined
Create (Composer markets, RFQ, webhook config)	fair-use	10× fair-use	contract-defined
WebSocket concurrent connections	fair-use	10× fair-use	dedicated stream, contract-defined

Fair-use quotas are published in X-RateLimit-Limit on every response and version-bumped with 90-day deprecation notice — hard-coding numbers into client code is discouraged; use the header instead.

Response headers

Every response includes:

X-RateLimit-Limit:     60
X-RateLimit-Remaining: 47
X-RateLimit-Reset:     1786190460  (UNIX seconds when window resets)

On 429 RATE_LIMITED you additionally get:

Retry-After: 13    (seconds until you can retry)

Upgrading tier

Builders routing volume through their apps, MMs, and institutional partners can upgrade to higher tiers:

Tier	Multiplier	Requires
Self-serve (sandbox)	1× w/ $10 order + $100/day notional caps	Anonymous mint via POST /v1/sandbox-keys
Read-live	1× baseline	Auto-graduated at $100 30-day attributed volume from self_serve — no application needed
Genesis (first 20 cohort)	10× baseline, custom bursts negotiable	Apply via POST /v1/apply with track=genesis — human review
Partner / Institutional	up to 20× or fully negotiable per contract	Signed partner / MM agreement with committed spread or volume
Load Test	ephemeral (up to 50×)	Submit a Request access ticket with test plan

Priority lanes

Sandbox and production traffic run on independent lanes with separate SLOs — sandbox load-testing never impacts read_live / trade_capped / trade_full / genesis / partner trade latency. WS event circuit_breaker fires if any lane enters degraded mode, letting well-behaved clients back off automatically.

Unauthenticated reads

Public unauthenticated reads (no key) are IP-throttled for anti-abuse. Mint an API key via POST /v1/sandbox-keys in 30 seconds to raise your quota to the self_serve baseline immediately, no email required.

Backoff code (Python)

import time, requests

def retry_with_backoff(fn, max_tries=5):
    for attempt in range(max_tries):
        r = fn()
        if r.status_code == 429:
            wait = int(r.headers.get("Retry-After", 2 ** attempt))
            time.sleep(wait)
            continue
        return r
    raise RuntimeError("max retries exhausted")

r = retry_with_backoff(lambda: requests.get(
    "https://api.predictasiax.com/v1/markets",
    headers={"X-Api-Key": "sk_live_ABC123"}
))
print(r.json())

WebSocket message throttle

Server-to-client messages have no rate cap. Client-to-server control frames (SUBSCRIBE / UNSUBSCRIBE / AUTH / LOCALE) are gently throttled to prevent misbehaving clients from exhausting a shared connection — batch multiple subscriptions into a single SUBSCRIBE call:

# Good:
{"method":"SUBSCRIBE","params":["fast_tick","signals","orderbook","account"]}

# Bad (4 messages instead of 1):
{"method":"SUBSCRIBE","params":["fast_tick"]}
{"method":"SUBSCRIBE","params":["signals"]}
{"method":"SUBSCRIBE","params":["orderbook"]}
{"method":"SUBSCRIBE","params":["account"]}