tool.hubspot_integration.5ebd51c72e954598

Use this skill when the OpenART registry selects tool.hubspot_integration.5ebd51c72e954598 for the current task.

HubSpot Integration Patterns

Build reliable HubSpot ↔ Airtable sync workflows. This skill covers API selection, rate limits, sync architecture, and n8n implementation.

Quick Reference

Task	API	Endpoint Pattern
CRUD objects	v3	POST /crm/v3/objects/{type}
Search/filter	v3	POST /crm/v3/objects/{type}/search
Batch upsert	v3	POST /crm/v3/objects/{type}/batch/upsert
Associations (all)	v4	POST /crm/v4/associations/{from}/{to}/batch/read
Labeled associations	v4	PUT /crm/v4/objects/{type}/{id}/associations/{toType}/{toId}

API Selection

Use v3 for Objects + Search

/crm/v3/objects/contacts          # CRUD
/crm/v3/objects/contacts/search   # Filter + paginate
/crm/v3/objects/contacts/batch/upsert  # Bulk writes

Use v4 for Associations

v3 associations only return the primary associated company. v4 returns all associations with labels.

/crm/v4/associations/contacts/companies/batch/read  # Batch read (1000 inputs)
/crm/v4/objects/contact/{id}/associations/company/{companyId}  # Labeled association

Authentication

Use Private App tokens (Bearer auth). API keys are deprecated.

Authorization: Bearer {ACCESS_TOKEN}

Required scopes for sync:

• crm.objects.contacts.read / .write
• crm.objects.companies.read / .write
• crm.objects.deals.read / .write
• crm.schemas.custom.read

Rate Limits

Tiered Limits (Private Apps)

Tier	Burst (per 10s)	Daily
Free/Starter	100	250,000
Professional	190	625,000
Enterprise	190	1,000,000

Search API: Separate Limit

5 requests/second per token — no rate limit headers returned.

Associations API (v4)

• Burst: 100-150/10s depending on tier
• Batch read: 1,000 inputs max
• Batch create: 2,000 inputs max

Rate Limit Headers

X-HubSpot-RateLimit-Remaining      # Requests left in 10s window
X-HubSpot-RateLimit-Daily-Remaining # Requests left today
Retry-After                        # Seconds to wait after 429

Backoff Strategy:

1. If Retry-After header exists, wait that duration
2. Otherwise, if X-HubSpot-RateLimit-Remaining < 10, wait 3 seconds
3. On 429 without headers, wait 10 seconds (TEN_SECONDLY_ROLLING policy)

Sync Architecture

Recommended: Hybrid Pattern

• Webhooks for immediacy (deletions, merges, critical updates)
• Polling with watermark for completeness (incremental sync)

Watermark Incremental Sync (Read Path)

// Store watermark: last_sync_timestamp (epoch ms)
{
  "filterGroups": [{
    "filters": [{
      "propertyName": "lastmodifieddate",  // Note: NOT hs_lastmodifieddate for contacts
      "operator": "GT",
      "value": "1674758400000"
    }]
  }],
  "sorts": [{ "propertyName": "lastmodifieddate", "direction": "ASCENDING" }],
  "limit": 200,
  "after": "cursor_from_previous_page",
  "properties": ["email", "firstname", "lastname", "hs_object_id"]
}

Critical: Use overlap window (watermark minus 2-5 minutes) to avoid missing records with identical timestamps.

Write Path (Airtable → HubSpot)

Use batch upsert with idProperty for idempotent writes:

POST /crm/v3/objects/contacts/batch/upsert?idProperty=email
{
  "inputs": [
    { "id": "user@example.com", "properties": { "firstname": "James" } }
  ]
}

Avoiding N+1 Queries

Bad: Fetch 100 contacts, then 100 individual association calls = 101 requests

Good: Batch read associations:

POST /crm/v4/associations/contacts/companies/batch/read
{ "inputs": [{ "id": "101" }, { "id": "102" }, ...] }  // up to 1000

Bi-Directional Loop Prevention

Prevent infinite sync loops with:

1. Field-level diffing: Only write if incoming value differs from existing
2. "Last Updated By" filter: Exclude changes made by integration user

   hs_updated_by_user_id NOT_EQ {integration_user_id}
   

Reference Files

• references/gotchas.md — Merge handling, deletions, property names, data formats
• references/n8n-patterns.md — Pagination loops, webhook patterns, rate limiting
• references/api-reference.md — Code examples for search, batch ops, associations