generated: '2026-09-14'
method: searched
source: >-
  https://github.com/boozallen/strands-base-agent/blob/develop/docs/foundry/guides/http-api.md and
  .../docs/foundry/configuration/environment-variables.md — read 2026-09-14
limit_count: 0
rate_limits: []
headers: []
exhaustion_status: null
note: >-
  No rate limits are published and none would be meaningful: the only Booz Allen service with a
  documented HTTP contract is self-hosted by the adopter, who sets their own capacity. No
  RateLimit-*/X-RateLimit-*/Retry-After headers are documented and 429 does not appear among the
  statuses the error middleware returns (400 / 404 / 422 / 500). An agent calling a deployed fork
  gets no runtime throttling signal at all.
input_bounds:
  note: >-
    What the baseline DOES enforce is per-request input bounds — the nearest published thing to a
    quota, and the values a client must respect to avoid a 422.
  limits:
    - {field: query, bound: '1-8192 characters, non-whitespace', config_default: 'max_query_length 2000'}
    - {field: session_id, bound: '8-128 chars [A-Za-z0-9_-] in body; 3-40 chars in path'}
    - {field: context, bound: '<= 32 keys, <= 16 KiB serialized'}
    - {field: max_results, bound: '1-100 (default 10)'}
    - {field: similarity_threshold, bound: '0.0-1.0 (default 0.7)'}
    - {field: limit, bound: '1-100 on chat-history listings'}
    - {field: offset, bound: '>= 0'}
    - {field: max_response_time_ms, bound: 'default 30000'}
  timeouts:
    - {name: AWS_READ_TIMEOUT, default: 900, unit: seconds, scope: Bedrock read}
    - {name: AWS_CONNECT_TIMEOUT, default: 60, unit: seconds, scope: Bedrock connect}
upstream_note: >-
  Real throughput limits on a deployed agent come from the model provider it is pointed at (AWS
  Bedrock by default), not from Booz Allen.