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: AlphaSense
providerId: alphasense
created: '2026-05-23'
modified: '2026-05-23'
reconciled: false
tags:
  - Market Intelligence
  - Search
  - Generative AI
  - Agents
  - GraphQL
  - Rate Limiting
  - Credits
description: >-
  AlphaSense does not publish numeric per-second or per-minute rate limits for
  the Agent API or Utility APIs in the public developer portal. The
  developer.alpha-sense.com documentation references a "Credits & Rate Limits"
  section that governs GenSearch and Deep Research consumption, but the credit
  pricing, monthly allowance, and throttling ceiling are negotiated per
  customer contract. Standard HTTP 4xx/5xx response codes are applied; 429 is
  returned on credit-pool or burst exhaustion. The /auth bearer token has a
  finite lifetime and must be refreshed. This artifact stays reconciled:false
  until paired with a real customer contract.
sources:
  - https://developer.alpha-sense.com/api/getting-started
  - https://developer.alpha-sense.com/agent-api/quickstart
responseCodes:
  throttled: 429
  unauthorized: 401
  serverError: 5xx
limits:
  - name: GenSearch / Deep Research Credit Pool
    scope: customer
    metric: credit
    limit: contracted
    timeFrame: month
    notes: >-
      Generative calls (GenSearch in any mode, Deep Research, Workflow Agents
      invoking generative steps) draw from a contracted monthly credit pool.
      Exhausting the pool returns 429 and requires top-up or contract uplift.
  - name: GraphQL Burst (per-customer)
    scope: customer
    metric: requests
    limit: contracted
    timeFrame: minute
    notes: >-
      Per-customer burst throttling on the shared /gql endpoint. Numeric
      ceiling not published; behaviour returns 429 on excessive throughput.
  - name: Ingestion Throughput
    scope: customer
    metric: documents
    limit: contracted
    timeFrame: hour
    notes: >-
      Enterprise Intelligence ingestion runs against a contracted ingestion
      quota; large initial loads are typically scheduled with Solutions
      Engineering.
  - name: Bearer Token Lifetime
    scope: session
    metric: token
    limit: short-lived
    timeFrame: hour
    notes: >-
      /auth returns a bearer token with a finite expires_in. Clients must
      refresh the token (or re-authenticate) before expiry.
policies:
  - name: 429 Throttling
    description: >-
      AlphaSense returns 429 on burst exhaustion or credit-pool exhaustion.
      Implement exponential backoff with jitter and treat persistent 429s as a
      signal to renegotiate the credit pool.
  - name: Token Refresh
    description: >-
      Refresh the bearer token before expires_in elapses; do not cache
      indefinitely. Long-running agent workloads should hold a credential
      manager that re-auths against /auth.
  - name: Polling Cadence
    description: >-
      GenSearch / Deep Research conversations are polled until progress reaches
      1.0. Recommended cadence is a few seconds per poll; do not tight-loop.
  - name: Credit Accounting
    description: >-
      Deep Research is materially more expensive in credits than fast or auto
      GenSearch modes. Route the right mode for the question to control burn.
maintainers:
  - FN: Kin Lane
    email: kin@apievangelist.com