Known differences

HubSpot

A HubSpot portal that answers the CRM, marketing, CMS, email, automation and engagements APIs the official Python SDK calls, plus HubSpot's remote MCP paths.

Vendor reference: https://developers.hubspot.com/docs/api/overview ↗. Machine-readable: compat/hubspot.json.

Surfaces

Kind	Path	Notes
REST	`/crm/v3`	objects, search, batch, properties, pipelines; associations under `/crm/v4`; schemas under `/crm-object-schemas/v3`
REST	`/marketing/v3`	also `/cms/v3`, `/email/public/v1`, `/automation/v4`, `/engagements/v1`, `/account-info/v3`
OAuth	`/oauth`	private-app tokens or OAuth2 access tokens, and HubSpot's remote-MCP OAuth paths
Webhooks	`webhook subscriptions`	CRM events delivered to your endpoint
MCP	`/mcp`	also `/anthropic`, `/openai` and the bare origin, as `mcp.hubspot.com` serves them

API versions: CRM v3 and v4, Marketing v3, CMS v3, Automation v4, Engagements v1

Supported

The official Python SDK (hubspot-api-client) unmodified, with host pointed at the replica
CRM search with filterGroups, batch read, and list and get across object types
Associations with HubSpot's HUBSPOT_DEFINED type ids and portal-defined USER_DEFINED labels
X-HubSpot-RateLimit-* headers on every response, refilling per ten-second interval with a daily countdown
HubSpot's error envelope

Not supported

Numeric record ids for generated records (ids look like acct-0)

Known differences and test guidance

Scenario	Difference from HubSpot	In your tests
Record ids	Generated records use string ids such as `acct-0`, not numeric ids; created records get numeric ids from `9000001`. Engagements also answer to a numeric alias.	Treat ids as opaque strings; do not do arithmetic on them or parse `toObjectId` as a number.
Rate limits	Rate-limit headers are sent but the limit is not enforced by default.	Assert that your client reads the headers; inject `throttle` to test a 429.
Campaign metrics	Metrics follow the campaign's email engagement events; a campaign with no events falls back to a deterministic synthetic curve.	Assert on consistency between events and totals, not on specific values.
State between tests	Created records are kept in memory, visible to every read path, and dropped by a reset or state change.	Reset between tests rather than cleaning up by hand.

Fault injection

Key	What the client sees
`throttle`	429 `RATE_LIMITS` with policy `TEN_SECONDLY_ROLLING` and `Retry-After`
`fail_next`	The next N calls fail as `throttle` does, then clear
`quota_limit`	429 `RATE_LIMITS` with policy `DAILY` ("You have reached your daily limit.")
`error_rate`	500 `INTERNAL_ERROR` for the given fraction of calls
`latency_ms`	Fixed added latency on every data-plane request

Applies to every system

The hosted data plane is read-only: a write is refused with 403 writes_disabled. A SOQL, GraphQL or search POST is a read and is answered.
Faults are injected through POST /_admin/faults on a simulator you run yourself; POST /_admin/faults/reset clears them.
Rate limits do not happen on their own unless a page says so. Use fault injection to exercise a client's backoff.
Distributions come from aggregated metadata sketches of data Eon backs up; no customer records; all Era data is simulated.