specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: LVL LTD CO
providerId: lvlltd-com
generated: '2026-09-19'
method: searched
probe: true
created: '2026-09-19'
modified: '2026-09-19'
sources:
- https://lvlltd.com/docs/REFERENCE.md#rate-limits-retry-idempotency
- https://lvlltd.com/api/a2a (dev_patterns.patterns.production.rate_limit; jsonrpc_transport.headers + error_codes)
- live POST https://lvlltd.com/api/a2a on 2026-09-19 (X-RateLimit-Remaining: 119 observed)
- live GET https://lvlltd.com/api/catalog on 2026-09-19 (no rate-limit headers observed)
tags: [Rate Limiting, A2A, x402]
description: >-
  LVL LTD publishes one hard limit — the A2A JSON-RPC endpoint at ~120 requests per 60 seconds per
  isolate, signalled by an X-RateLimit-Remaining response header, HTTP 429 and JSON-RPC code -32005 —
  and says the REST skill-pay path has "no hard public rate limit on canary path; abuse may be throttled
  at edge" (Cloudflare). A live REST request carried no RateLimit-* headers; a live A2A request returned
  X-RateLimit-Remaining: 119, which corroborates the 120/60 s figure. The MCP endpoint documents no limit.
limit_count: 1
headers:
  remaining: X-RateLimit-Remaining
  requestId: X-Request-Id
  retryAfter: null
  note: Headers are emitted by the A2A endpoint only; REST responses expose cf-ray and no limit headers.
responseCodes:
  throttled: 429
  jsonrpc: -32005 RATE_LIMIT
limits:
- name: A2A JSON-RPC requests
  scope: per isolate (Cloudflare Worker instance)
  endpoint: https://lvlltd.com/api/a2a
  metric: requests
  limit: 120
  timeFrame: 60 seconds
  burst: null
  observed: 'X-RateLimit-Remaining: 119 after one request (2026-09-19)'
undocumented:
- surface: REST /api/* (pay, catalog, proof, shop ...)
  note: '"No hard public rate limit on canary path; abuse may be throttled at edge" — no number, no header. Terms section 6 forbids "abuse of rate limits".'
- surface: MCP /api/mcp
  note: No limit documented; the agent-memory skill mentions "rate limiting per wallet" as a feature of that paid skill only.
retry_guidance: 'Exponential backoff 2 s, 4 s, 8 s for transient 5xx and PAYMENT_VERIFICATION_FAILED (retry:true); do not retry terminal A2A tasks.'
spend_caps: 'Per-wallet spend caps via GET/POST /api/meter (MCP meter_budget) — "Caps prevent runaway agent spend"; paid_api_call is capped at $5 per call.'