generated: '2026-07-18'
method: searched
source: https://docs.corti.ai
description: >-
  Cross-cutting request/response semantics for the Corti API, captured from the
  authentication, errors, and API-reference documentation. No documented
  idempotency-key mechanism was found, so no Idempotency pointer is emitted.
authentication:
  style: oauth2-client-credentials
  header: 'Authorization: Bearer <jwt>'
  additional_required_header: 'Tenant-Name: <tenant>'
  token_ttl_seconds: 300
  ref: authentication/corti-authentication.yml
versioning:
  style: uri-path
  current: v2
  base_url: https://api.{environment}.corti.app/v2
  environments: [eu, us, beta-eu]
tenancy:
  header: Tenant-Name
  default: base
  note: Enterprise setups use a custom tenant/realm name.
rate_limiting:
  signal: HTTP 429 Too Many Requests
  client_behavior: SDKs auto-retry with exponential backoff (default maxRetries 2, overridable per request)
  ref: errors/corti-error-codes.yml
error_envelope:
  rest: JSON body; 401 details in WWW-Authenticate header
  websocket: CONFIG_* status code with reason field
  ref: errors/corti-error-codes.yml
idempotency:
  supported: false
  note: No idempotency-key header documented as of capture.
realtime:
  protocol: WebSocket (wss)
  specs: [asyncapi/corti-transcribe-asyncapi.json, asyncapi/corti-stream-asyncapi.json]
  note: Bidirectional streaming for stateless dictation (transcribe) and ambient documentation (stream).
model_compatibility:
  openai: Chat Completions, Responses, Completions, Embeddings, Models endpoints
  anthropic: Messages endpoint
  note: Corti Models are OpenAI/Anthropic-compatible, EU-hosted.