generated: '2026-09-19'
method: searched
source: https://cymetica.com/api-docs (Rate Limits, API Key Management, Quick Start tiers) + https://cymetica.com/llms-full.txt
  + MCP tools/call get_trust_ladder (public) 2026-09-19 + https://cymetica.com/.well-known/agent-card.json (communication_guide)
  + observed response headers
limit_count: 13
summary: 'Limits are published in three independent ladders — REST general requests per API key, REST order placement
  per key tier, and MCP calls per agent trust level — plus two per-endpoint limits. The runtime signal is the X-RateLimit-Limit
  / X-RateLimit-Remaining / X-RateLimit-Reset trio with HTTP 429 on exhaustion. Observed live: an anonymous GET
  /api/v1/markets returned x-ratelimit-limit: 2000, which matches none of the documented numbers (100 authenticated
  / 20 unauthenticated), so the deployed general budget is looser than the docs state.'
headers:
  limit: X-RateLimit-Limit
  remaining: X-RateLimit-Remaining
  reset: X-RateLimit-Reset (Unix timestamp)
  retry_after: null
  exhaustion_status: 429
  observed:
  - url: https://cymetica.com/api/v1/markets
    auth: none
    x-ratelimit-limit: 2000
    x-ratelimit-remaining: 1999
    x-ratelimit-reset: 1789872840
  - url: https://cymetica.com/openapi.json
    auth: none
    x-ratelimit-limit: 2000
limits:
- scope: per-api-key
  tier: Standard
  class: general API requests
  limit: 100
  window: 1 minute
  source: https://cymetica.com/api-docs
- scope: per-api-key
  tier: Premium
  class: general API requests
  limit: 1000
  window: 1 minute
  source: https://cymetica.com/api-docs
- scope: per-api-key
  tier: Admin (internal)
  class: general API requests
  limit: 10000
  window: 1 minute
  source: https://cymetica.com/api-docs
- scope: per-ip
  tier: unauthenticated
  class: general API requests
  limit: 20
  window: 1 minute
  source: https://cymetica.com/llms-full.txt ("100 req/min (authenticated), 20 req/min (unauthenticated)")
- scope: per-api-key
  tier: STANDARD
  class: order placement
  limit: 3000
  window: 1 minute
  source: https://cymetica.com/api-docs
- scope: per-api-key
  tier: PREMIUM
  class: order placement
  limit: 12000
  window: 1 minute
  source: https://cymetica.com/api-docs
- scope: per-api-key
  tier: MARKET_MAKER
  class: order placement
  limit: 60000
  window: 1 minute
  note: DMM agreement required
  source: https://cymetica.com/api-docs
- scope: per-agent (MCP)
  tier: unknown (rank 0)
  limit: 10
  window: 1 minute
  also:
    per_hour: 100
    per_day: 500
    concurrent_connections: 1
  source: MCP get_trust_ladder
- scope: per-agent (MCP)
  tier: recognized (rank 1)
  limit: 30
  window: 1 minute
  also:
    per_hour: 500
    per_day: 5000
    concurrent_connections: 3
  source: MCP get_trust_ladder
- scope: per-agent (MCP)
  tier: trusted (rank 2)
  limit: 60
  window: 1 minute
  also:
    per_hour: 2000
    per_day: 20000
    concurrent_connections: 5
  source: MCP get_trust_ladder
- scope: per-agent (MCP)
  tier: allied (rank 3)
  limit: 200
  window: 1 minute
  also:
    per_hour: 10000
    per_day: null
    concurrent_connections: 20
  source: MCP get_trust_ladder
- scope: per-ip
  endpoint: POST /api/v1/bounty/qa-agent/ask
  limit: 20
  window: 1 minute
  source: https://cymetica.com/.well-known/agent-card.json communication_guide.sync_endpoint.rate_limit
- scope: per-agent
  endpoint: POST /mcp/v1/register response
  limit: 60
  window: 1 minute
  note: 'Docs example response `rate_limits.requests_per_minute: 60` for a freshly registered agent at trust_level
    recognized — an example value, kept separate from the ladder above.'
  source: https://cymetica.com/api-docs
penalties:
  api_violations: -50 trust points per rate-limit or policy violation (get_trust_ladder points_per_action.api_violations)
gaps:
- Retry-After is not documented or observed.
- The general-request numbers disagree across sources (docs 100/20 vs observed header 2000).
- The SDK advertises client-side "Rate Limiting" but publishes no algorithm.