specification: API Commons Rate Limits
specificationVersion: '0.1'
provider: Landbase
providerId: landbase
generated: '2026-08-23'
method: searched
created: '2026-08-23'
modified: '2026-08-23'
source: >-
  https://cli.landbase.com/latest/skills/tools/landbase-search/SKILL.md (the provider's own published
  Agent Skill, section "Concurrency Limits"),
  https://www.landbase.com/docs/reference/error-codes,
  https://www.landbase.com/docs/what-can-you-do-with-the-cli (contact-enrich batching guidance),
  https://www.landbase.com/docs/reference/contacts-import (polling guidance).
description: >-
  Landbase has no rate-limit page in its documentation, but it does publish real numbers — they are
  scattered across the CLI guide, the contacts-import reference and, most usefully, the Agent Skill
  Landbase ships to Claude Code and Codex, which states the concurrency ceiling and the client's
  retry behaviour outright. The runtime signal is a 429 with Retry-After, surfaced to the caller as
  a RATE_LIMITED error carrying meta.retry_after_seconds. Quota, as opposed to rate, is enforced
  commercially in credits rather than in requests — see plans/landbase-plans-pricing.yml.
docs: https://www.landbase.com/docs/reference/error-codes
headers:
  retry_after: Retry-After
  ratelimit_limit: null
  ratelimit_remaining: null
  ratelimit_reset: null
  request_id: null
  note: >-
    No X-RateLimit-* or RFC 9239-style RateLimit-* headers are documented, so a caller cannot see
    how much budget is left before it runs out — only that it has run out. Retry-After on the 429 is
    the entire signal.
responseCodes:
  throttled: 429
  edge_timeout: 524
error_envelope:
  code: RATE_LIMITED
  exit_code: 2
  meta_field: retry_after_seconds
  example: '{"error":{"code":"RATE_LIMITED","message":"Too many requests. Please wait before retrying.","meta":{"retry_after_seconds":30}}}'
client_retry:
  attempts: 3
  backoff: [5s, 10s, 20s]
  honours_retry_after: true
  source: https://cli.landbase.com/latest/skills/tools/landbase-search/SKILL.md
limits:
  - name: Concurrent agent runs
    scope: account
    metric: concurrent_runs
    limit: 10
    timeFrame: concurrent
    exhaustion: HTTP 429 / RATE_LIMITED
    note: >-
      "Max 10 concurrent runs per account. HTTP 429 = rate limited. Fan-out beyond 10 workers
      produces no speedup and burns retries." Published by Landbase in its landbase-search Agent
      Skill, not in the documentation.
  - name: Contact-enrichment batch size
    scope: request
    metric: leads_per_batch
    limit: 100
    timeFrame: request
    note: Published guidance — "keep each batch at or below 100 leads".
  - name: Contact-enrichment batch submission rate
    scope: account
    metric: batches_per_minute
    limit: 10
    timeFrame: minute
    note: Published guidance — "avoid submitting more than about 10 batches per minute per account". Stated as approximate.
  - name: Polling cadence
    scope: account
    metric: guidance
    limit: null
    timeFrame: null
    note: >-
      "Avoid sub-second polling — it won't return results faster and may trip rate limits." Landbase
      publishes a backoff schedule (5s, 10s, 20s cap) rather than a numeric poll ceiling.
result_ceilings:
  - {tool: Search Companies, limit: 10000, unit: companies per query}
  - {tool: Expand Lookalikes, limit: 20000, unit: scored companies}
  - {tool: Scout Lookalikes, limit: 25, unit: companies (default)}
timeouts:
  search_autopoll_default: 300s (overridable with --timeout)
  contact_enrich_wait_default: ~600s
  contacts_import_wait: 30 minutes overall, 5s/10s/20s backoff
limit_count: 3
gaps:
  - No published rate-limit reference page; the numbers must be assembled from four different documents.
  - No per-plan or per-tier limits are published, and no way to see remaining budget before exhaustion.
  - >-
    The most precise numbers Landbase publishes live in an Agent Skill rather than in its
    documentation — machine-readable to an agent that installs the CLI, invisible to a human reading
    the docs site.