Configuration
Create a key in your Mailchimp account’s Extras → API keys settings. The
server is the suffix after the last hyphen in the key. Requests use
https://<server>.api.mailchimp.com/3.0. Access depends on the key owner’s role
and the account’s plan. See Mailchimp’s authentication guide.
Connection testing reads the API root account endpoint.
Resources
Lists (audiences), members, campaigns, reports, and merge-field definitions are selected by default. All other resources are optional, including commerce, automations, SMS campaigns, account administration, and per-member detail reads. Selecting a child resource also reads its parents, but only selected resources are written to the destination.members includes contact status, merge fields, interests, tags, marketing
permissions, and statistics. campaign_content reads HTML and plain text.
Reports include delivery, engagement, and commerce summaries, with separate
optional resources for recipients, opens, clicks, and email activity.
Commerce IDs are scoped to their store. Line items, product variants, images,
and promo codes have separate paginated resources in addition to nested JSON
returned by their parents. File and account-export resources contain metadata
and URLs; they do not download files or create export jobs.
IDs are strings. Nested objects and arrays remain JSON. Most date fields remain
strings so that empty provider values are preserved; the incremental member
cursor is a timestamp with a time zone. Each resource has a raw column for
returned fields that are not projected separately.
The connector does not include Transactional/Mandrill, Open Commerce, beta
omni-channel audiences, landing pages, surveys, Facebook ads, conversations,
member goals, or batch-operation jobs. Alternate lookup/search endpoints and
campaign send checks are not separate resources. It does not send campaigns,
modify subscriptions, or receive webhook events.
Modes
All resources support full reads.members also supports incremental reads
using last_changed with the since_last_changed filter and a 60-second overlap.
Each audience is listed on every run. Member detail resources use a separate
full member index, so selecting incremental members does not hide older contacts
whose notes, tags, or activity need to be read.
The member cursor tracks changes to contact information. It is not a change feed
for every engagement statistic or related resource. Use periodic full reads to
refresh those values and reconcile archived or deleted members. Incremental
reads do not remove destination rows. Campaigns, reports, segments, and commerce
remain full-read because creation or send dates do not capture later changes.
Use full replace for member_events, member_activity, member_activity_feed,
signup_forms, report_locations, and report_product_activity: these have no
stable primary key declared. Keyed resources support full upsert or full replace.
Behavior
- Auth: requests use Bearer authentication with the API key. The credential inherits its owner’s role and is revoked when that user loses account access.
- Pagination: endpoints that document both
countandoffsetare read in pages of 100 until a short page. Other endpoints use their documented single response. Offset pagination is not a consistent snapshot during concurrent edits. - Dependencies: optional per-member reads can require one request per contact. Child reads use two concurrent requests. Hidden member indexing reads IDs only.
- Rate limiting: Mailchimp permits ten simultaneous connections. The engine
retries HTTP 429 responses and honors
Retry-After; it does not impose an account-wide connection cap shared with other clients. Select smaller resource groups if broad parallel reads exhaust that allowance. Permission errors remain failures rather than becoming empty results. - History:
member_activityreturns the last 50 events per member, not an unlimited event archive. Report arrays contain the activity returned by Mailchimp, with no promise of events outside its retention or tracking settings. - Access: optional resources can require account features, historical classic automations, or additional permissions. Static resource discovery does not validate those permissions. Endpoint failures remain visible.