attach_group_contacts
ChatGPTAttach up to 500 existing contact IDs to a static segment. Use payload={"contactIds":[123,456]} without profile or event fields. The upstream also supports externalCustomerIds, mutually exclusive with contactIds. A success acknowledgement can hide missing IDs; verify the requested memberships with get_group_contacts before claiming completion.
bulk_upsert_contacts
ChatGPTCreate or update many contacts; optional events and field changes can trigger workflows.
count_segment_contacts
ChatGPTCount contacts matching an unsaved dynamic-segment schema. Pass Segment Designer SDK finish().model UI-mode JSON as schema, preserving include, exclude, and parameters. No segment is saved and no contact rows are returned. For an existing segment ID, use get_segment_contacts instead. This tool applies exactly the supplied conditions, without adding an active-contact filter. Starts one calculation and waits up to 600 seconds (10 minutes), including account lookup; returns totalCount only for COMPLETED (including zero), with mediaChannels. Available to authenticated Reteno, Yespo, and eSputnik users. Limited to 6 attempts per 60 seconds per organization, including failures. Polls consume no additional allowance. Do not automatically retry an unknown or timed-out calculation.
create_app_inbox_message
ChatGPTCreate a base app inbox message.
create_dynamic_segment
ChatGPTCreate a live dynamic segment from Segment Designer SDK UI-mode JSON. Pass the runner's .segment.json as schema, not its diagnostics or a DB-mode schema. The payload follows the eSNG create wizard: name, description, tags, contactsCount=0, and schema are sent together to the contacts gateway. The gateway owns validation and UI-to-DB conversion. Authenticated Reteno, Yespo, and eSputnik users can see and call this tool without an agent token or organization allowlist entry. A successful create is fetched by ID to confirm that the saved segment is retrievable. The backend may normalize the submitted UI schema, so the fetched schema is not compared with the request. If the response has no ID or read-back is unavailable, search by exact name before retrying: the segment may already have been created.
create_email_message
ChatGPTcreate_email_message: finalize uploaded HTML/CSS and create a saved email. Creates a Yespo/Reteno regular email or reusable template from a prepared upload. Requires payload.uploadSessionId and payload.subject; source HTML/CSS stay in the uploaded files, not tool arguments. Preparation alone does not persist an email. Returns the created identity and write/read-back evidence; inspect it before retrying an uncertain write. `template defaults to false (a regular email). Set it to true to create a reusable template. Pass the discriminator as this top-level argument, not inside payload. When payload.tags` is present it must be a string array; names are trimmed, blank entries removed, and exact duplicates collapsed while preserving order and case. Use payload.sender for an exact private sender value. Dynamic payload.from expressions also use this path. A regular email is created with a registered interface, then its sender is PATCHed and read back. With payload.sender, an optional static payload.from selects that temporary interface; otherwise the first available interface is used. Templates may omit from. The requested resource kind is preserved. A failed sender step returns the created ID and arguments for resume_email_sender_update; never create again. Saving a dynamic sender does not verify feature enablement, domain verification, or sending readiness.
create_email_shell
ChatGPTMint a from-scratch email SHELL for a JSON-native (structured) compose. Creates a minimal, esd-native placeholder email message via the OAuth `/api/v1 path, RECORDS it in the short-TTL shell registry, and creates its initial CRDT model through Stripo's internal listener. It returns {messageId, modelId, modelStatus, expiresAt} only after both resources are ready. The JSON compose flow then: get_email_model the shell, edit the downloaded model with the stripo SDK, prepare_email_model_upload(operation="create") (accepted ONLY because this message is a registry-bound shell), upload, then update_email_model -- whose create-mode gate refuses to persist unless the composed document adds content BEYOND the bare seed (the minimum-content gate). The shell is single-use (consumed on the successful persist) and swept from the registry by TTL if the compose is abandoned; an abandoned durable message is recognizable by still carrying only the bare unsubscribe-only seed content in its rawHtml (never a tag). name and subject are required message metadata. Later subject changes use update_email_metadata. Shell creation has no preheader argument: set model.metadata.preheader in the native JSON during compose and persist it with update_email_model. Verify the saved preheader in a fresh model and compiled HTML export. The legacy flat update_email_metadata(preheader=...) path does not guarantee persistence. tags is optional and written through verbatim (no machine tag added). sender is the org email interface value (e.g. '"Name" <addr@dom>') used as the REST from -- REQUIRED by the create verb; when omitted the FIRST interface returned by GET /api/v1/interfaces/email is used. template=true mints the shell as a reusable email template; the default false mints a regular email message. Custom fonts are optional. fonts declares at most four Google Fonts families and the exact static weights to preload before Stripo creates the model; italic=true adds italic faces as well as normal faces. font_families assigns a declared family to body, headings, and/or buttons. Common web-safe families (Arial, Helvetica, Georgia, Times New Roman, Verdana, Tahoma, Trebuchet MS, and Courier New) may be assigned without declaring them in fonts. Arbitrary stylesheet URLs are intentionally unsupported: generated font resources are limited to fonts.googleapis.com and fonts.gstatic.com. Loading custom fonts can access these external Google Fonts resources. Seed contract: the packaged seed is the owner's editor-captured "unsubscribe-only" starter, minted BYTE-VERBATIM (zero templating). Its entire content is legitimate-to-keep, so there is no placeholder to replace and no per-mint marker to gate on; the finalize's minimum-content gate is what proves a real (content-bearing) compose landed. Recovery contract: after an ambiguous REST create that returns no usable id, list messages by the exact name` and adopt the existing match before attempting another mint. A later model-create failure rolls the identified REST shell back automatically.
create_event
ChatGPTCreate a current v3 event; it may trigger automation and create a new event type. params is a JSON object, for example {"key":"value"}; omit it or use null when no parameters are needed. Legacy arrays of name/value pairs are not v3 input. No pre-existing event type is required by this tool.
create_mobile_push_message
ChatGPTCreate a base mobile push message.
create_past_events
ChatGPTUpload historical events.
create_sms_message
ChatGPTCreate a base SMS message.
create_workflow
ChatGPTCreate a new inactive workflow shell for the authenticated Reteno organization. Available to authenticated Reteno, Yespo, and eSputnik users. No agent token or organization allowlist entry is required. The created workflow is paused and contains a Start block. Call get_workflow_export to obtain modelHash, then update_workflow with SDK code and expected_model_hash to build the workflow. Retain the returned workflowId; reconcile an ambiguous create before retrying.
delete_app_inbox_message
ChatGPTDelete a base app inbox message.
delete_app_inbox_message_translation
ChatGPTDelete a translated app inbox message variant.
delete_broadcast
ChatGPTDelete a broadcast by ID.
delete_contact
ChatGPTDelete a contact by ID.
delete_contact_by_external_customer_id
ChatGPTDelete a contact by external customer ID.
delete_email_message
ChatGPTDelete an email message by ID.
delete_email_message_translation
ChatGPTDelete a translated variant and reconcile its Reply-To recovery. Confirmed deletion retires old operations. Recreating the language requires a fresh upload and a new replyToOperationId.
delete_mobile_push_message
ChatGPTDelete a base mobile push message.
delete_mobile_push_message_translation
ChatGPTDelete a translated mobile push message variant.
delete_sms_message
ChatGPTDelete a base SMS message.
delete_sms_message_translation
ChatGPTDelete a translated SMS message variant.
detach_group_contacts
ChatGPTRemove up to 500 contact IDs from a static segment. Use payload={"contactIds":[123,456]}. The upstream also supports externalCustomerIds, mutually exclusive with contactIds. An empty payload removes ALL members; never use it for a selected-ID operation. Verify resulting membership with get_group_contacts.
edit_image
ChatGPTEdit ONE image for Reteno email and campaign design, including text and masks. Rate limits apply. This tool does not initiate a purchase or payment. Maximum 10 files TOTAL: base image + references + optional mask. Each input uses exactly one HTTPS url or uploadSessionId. The PNG mask must have alpha and the same dimensions as the base; transparent areas guide edits. Mask boundaries and generated text are model guidance, not exact pixel/text guarantees. textPolicy is preserve (unless explicitly asked to change text), replace, or none (remove text). Default delivery size is the source size, within 1536px/1,572,864 pixels. preserveAlpha defaults true for transparent sources; it restores the original alpha and requires the original dimensions plus PNG/WebP. TinyPNG is on by default; exact source alpha is restored after compression and the final file is checked against format, dimensions, and size limits. Returns a jobId immediately. Poll get_image_job. Retry the identical call with the same idempotencyKey to recover saved processing without repeating generation. Wait for error.retryAfterSeconds when present. The result is hosted; no email is modified.
generate_image
ChatGPTCreate ONE new image for Reteno email and campaign design, with up to 10 references. Rate limits apply. This tool does not initiate a purchase or payment. Inputs use exactly one HTTPS url or uploadSessionId. References guide a new composition; their layout, text and alpha are not inherited. size is the final WIDTHxHEIGHT (default 1024x1024, no 'source'). Default limits: 1536px edge, 1,572,864 pixels, 3 MiB/file. Smaller delivery dimensions are resized from a supported provider resolution. TinyPNG compression is on by default. No email content is changed. Returns a jobId immediately. Poll get_image_job for the hosted image. Use a unique idempotencyKey for each intentional new generation. After a timeout or retryable processing failure, wait for error.retryAfterSeconds when present, then repeat the IDENTICAL call with the SAME key; saved generation is never repeated. A new key can start a separate generation. No filesystem paths or inline image bytes.
get_account_info
ChatGPTGet Reteno account information.
get_addressbooks
ChatGPTGet account address books.
get_app_inbox_message
ChatGPTGet an app inbox message by ID.
get_brandkit
ChatGPTGet the saved Brand Kit for the authenticated organization. The normalized document is returned exactly as stored. An empty stored list is a value the server holds; only an entirely omitted language list is filled from its twin. This is the temporary legacy alias of get_business_profile.
get_broadcast
ChatGPTRead a broadcast. Private scheduled reads include persisted timezone, batching, and verification fields. Never infer contact-local scheduling from startDate alone. Legacy/historical reads do not verify that setting.
get_business_profile
ChatGPTGet the saved Business profile for the authenticated organization.
get_contact
ChatGPTGet a contact by ID.
get_contact_analytics
ChatGPTRead native contact-base dashboards, status, growth, tracked groups, RFM and cohort reports. Growth/cohort/group reports use from_day/till_day calendar dates and native date_unit. Activity uses date_from/date_to (date or ISO datetime), resolved in the organization timezone. Disabled reports remain disabled; no tracking or feature is activated. Returns a signed JSON report to download and inspect. contact_history requires a contact_id and both date bounds; it reads one bounded native interaction page with channel identities including Widget. Grouped rows and unsafe native cursor ties prevent an exhaustive-history claim. It cannot replace account-wide historical cohort inputs.
get_contact_devices
ChatGPTGet mobile/web push devices for a contact by ID from the contact-details UI gateway. The public contact API omits the application device list. This mirrors the frontend contact-details resolver, which reads /contacts-gateway/rest/v1/contact/{contactId} and renders contact.applications[].devices.
get_contact_emails_by_ids
ChatGPTReturn email addresses for 1–100 known contact IDs in the connected account. Use when contact IDs are already available and their email addresses are needed. Returns only matching contactId/email pairs, always as a results list. This read-only lookup does not search by email or modify contacts.
get_contact_id_by_token
ChatGPTGet a contact ID by mobile push token.
get_contact_in_app_message_statuses
ChatGPTGet READ and CLICKED activity for one contact and true In-App message. Provide exactly one identifier supported by the documented contacts activity endpoint: email, phone, or Mobile Push token. The tool asks separately for READ and CLICKED activity, filters mediaType=inapp and the requested message ID locally, follows the endpoint's opaque offset, and bounds both scanning and returned evidence. date_from and date_to use ISO date/date-time format and may span at most three months.
get_contact_message_history
ChatGPTGet a contact's message history from the v2 behavior tracking API.
get_contacts_activity_v2
ChatGPTGet contacts activity from the v2 behavior tracking API.
get_email_deliverability_setup
ChatGPTGet a read-only snapshot of email domain, sender, short-link, and unsubscribe setup. Use this tool to inspect authenticated organization setup for email deliverability. It reads private frontend configuration endpoints and never creates, updates, or deletes domains, SSL settings, or sender configuration. If a per-domain detail request fails, that domain includes an error while the rest of the snapshot is still returned.
get_email_message_export
ChatGPTGet export links for one regular email or template and its translations. The JSON artifact preserves Unicode text but necessarily represents markup as a JSON string. Use the separate HTML artifact for literal rendered-content checks and punctuation-sensitive verification. The public content and authoritative private Reply-To/sender metadata are read here and the artifacts are written to disk, so the returned links are a snapshot: they keep serving this content even after the message changes upstream. Re-run the tool to export the current state.
get_email_message_preview_png
ChatGPTRender Gmail-style PNG previews for a regular email or template. Renders the saved email HTML/CSS in a Gmail-like wrapper without reading or initializing a Stripo model. Works for HTML-only emails, including newly created onboarding messages. The response reports `previewSource: durable-compiled-html, modelHash: null, and modelFidelity.structuredModel: unverified; it does not verify model edits that have not been saved to HTML. mode accepts mobile (default), desktop, or both`.
get_email_message_view_link
ChatGPTGet the hosted preview link for a regular email or template.
get_email_model
ChatGPTGet an email's structured model manifest plus subject and preheader. Returns a small manifest -- a short-lived signed download URL for the {settings, stripes} model FILE plus modelHash, a fidelity summary, and `messageMetadata` containing the stored subject and effective Hidden Preheader. Metadata normally comes from the authenticated private content API. When that API reports the preheader as empty, null, or absent, the value is reconciled against the durable compiled export so unavailable metadata is never mislabeled as a genuine empty preheader. The model read uses Stripo's direct document-state API. A specific upstream 500 can require one native editor visit before retrying it. The model is NEVER in the tool result; download it to a local file via the proxy, then edit it with the stripo SDK. For rendered HTML/CSS artifacts use get_email_message_export instead. When fidelity.structuredModel is "empty" (blockCount 0 and moduleCount 0) the email is an HTML-only clone with no editable Stripo model -- there is nothing to edit structurally, so use get_email_message_export for its content. A container carrying only id and settings with no blocks array is valid: blocks is optional and a missing blocks array should be treated as an empty list.
get_email_model_schema
ChatGPTGet a signed download manifest for the current Stripo JSON model schema. Fetches the deployed Draft-07 schema on every call. Download the JSON file from downloadUrl to validate {settings, stripes} models locally; sha256 verifies the downloaded bytes. The schema is never returned inline. No message ID or editor session is needed.
get_event_types
ChatGPTList event types for the authenticated Reteno organization. By default, this returns only active event types. Set include_inactive=true to include inactive event types returned by Reteno. Use this tool to find valid event types by ID, exact event key, or text search before calling event analytics tools. It cannot return counts, trends, delivery metrics, failure diagnostics, or other analytics results. Returns event-type metadata only.
get_events_analytics
ChatGPTExport aggregated event counts for the authenticated Reteno organization. Use this tool to summarize live event volume by event type with optional date/datetime and event-type filters. date_from/date_to accept YYYY-MM-DD or ISO datetimes; date-only and timezone-less datetime values are interpreted in the organization timezone. Requested ranges may span at most 185 days. Date-bucket splits and source filters are not available from the live API source. It cannot return raw event rows, user-level breakdowns, or delivery-diagnostic detail such as recipient domains or bounce reasons. Same-day ("today") analytics are available and may change as new events arrive. Returns a temporary CSV download manifest with the resolved time range, applied filters, row count, and freshness metadata.
get_group_contacts
ChatGPTSearch contacts in a segment by segment ID.
get_image_job
ChatGPTRead image job status or its hosted result in the authenticated organization. This tool never starts/resumes generation, compression, or upload. Poll at pollAfterSeconds while running. For retryable failures or interrupted processing, wait for error.retryAfterSeconds when present, then repeat the original generate_image/edit_image call with its original arguments and idempotencyKey. Unknown provider outcomes are never regenerated automatically.
get_import_status
ChatGPTCheck the status of a previously started asynchronous contact import by session ID.
get_in_app_analytics
ChatGPTRead In-App message performance: sends, engagement, conversions and revenue. Use discovery to find campaigns, detail/progress for campaign reports, and timed or detailed reports for breakdowns. App Inbox is a separate channel. Contact categories SEND, READ and CLICKED show current report membership, not verified historical membership at a past cutoff. Dates use the account timezone; omit both for all-period discovery/detail/progress. Timed reports require both dates; dimension discovery does not accept date filters. Returns a signed JSON report download link; no campaigns are changed.
get_message_status
ChatGPTGet delivery status for a single sent message.
get_messaging_analytics
ChatGPTExport aggregated messaging performance for the authenticated organization. Use this tool to compare live messaging metrics grouped by organization, workflow, or message, with optional date/datetime and supported filters. For whole-workflow analytics use group_by="workflow" and workflow_ids; omit media_types and message_ids. Workflow rows cover all native channels, with blank media_type, and include launches. Channel-specific workflow totals are unavailable. Organization and workflow rows include reach and frequency; spam, bounced, and unsubscribed are available only on message rows. When media_types is omitted, organization/message analytics default to email and mobilepush; workflow analytics cover all channels. Multi-channel requests are split into one live API request per channel for callout-backed message rows. date_from/date_to accept YYYY-MM-DD or ISO datetimes; date-only and timezone-less datetime values are interpreted in the organization timezone. Requested ranges may span at most 185 days. The finest available grouping level is callout/message; date-bucket splits are not available from the live API source. It cannot show recipient-domain splits, mailbox-provider or ISP splits, bounce reasons, recipient-level delivery errors, or raw event or log data. Same-day ("today") analytics are available and may change as new events arrive. Returns a temporary CSV download manifest with the resolved time range, applied filters, row count, and freshness metadata. Workflow rows require a launch in the requested period; an absent row does not prove zero activity. For daily reports request each local day separately; retain the native full-period result rather than summing daily unique counts. Accepted media_types for Reteno are email, sms, mobilepush, appinbox, and webpush. conversion_event selects event conversions by EventType ID or an exact case-insensitive event key/name. When supplied, the CSV includes conversions, and order_by may sort by conversions.
get_mobile_push_message
ChatGPTGet a mobile push message by ID.
get_mobile_push_token_activation
ChatGPTGet mobile push token activation state.
get_organisation_info
ChatGPTGet organization-scoped inventory data for Workflow Designer style resolution. Each optional category argument must be an object shaped like {"enabled": true, "query": "optional search text", "id": "optional exact id"}. Set enabled only for the categories needed in the current step; query narrows the text search when the underlying endpoint supports it. id may be a string, integer, or list of strings/integers, and takes precedence over query for exact ID resolution. For contact_fields, optional include_groups: true also returns data.contact_field_groups with native group IDs, names and keys, including empty groups. Group choices are unfiltered by query/id; existing contact_fields summaries keep their normal filtering. Supported categories: contact_fields, email_messages, sms_messages, viber_messages, telegram_messages, mobile_push_messages, web_push_messages, app_inbox_messages, tags, segments, webhooks, events. The response data always includes the organization timezone and its current GMT offset in minutes. Returns {"success": true, "data": ...} when every requested category succeeds. If one or more categories fail, the tool returns {"success": false, "data": <partial>, "errors": {...}} so callers can still use the categories that were retrieved successfully.
get_recommendation_analytics
ChatGPTRead website/mobile recommendation metrics and attributed purchase rows. IDs must belong to the selected platform's live discovery. Date bounds use the organization timezone and are sent as UTC ISO timestamps with millisecond precision, matching the UI. Purchases are product rows, not unique orders; complete=true traverses pages and fails rather than truncating at a cap. Purchases default to the UI's 10 rows/page; the default 100-page cap covers up to 1,000 product rows. Purchase pages can still fail upstream; complete traversal then fails without partial totals. Returns native metrics/currency in a signed JSON report. Completed traversal does not prove snapshot consistency or financial attribution definitions.
get_segment
ChatGPTRead org-info segment metadata by ID and optionally include its dynamic rule schema.
get_segment_contacts
ChatGPTReturn the active-contact count and optionally one page of people in a saved segment. This uses the same GROUP-by-ID contact-schema request as the eSNG segment contacts drawer. Callers pass a saved segment ID, never a segment schema. Set include_contacts=false for a count-only response. Account lookup and contact retrieval share a 600-second (10-minute) deadline.
get_segment_definition
ChatGPTRead live dynamic-segment condition definitions. Compatibility alias for list_segment_definitions(...).
get_segment_schema
ChatGPTRead a live dynamic-segment schema; absolute dates also include ISO normalizedValue metadata.
get_sms_message
ChatGPTGet SMS message content, title, sender, and text by internal SMS message ID.
get_widget_analytics
ChatGPTRead Web Widget campaign reports or Claspo/form performance and new contacts. Reports → Web Widget: campaign_list returns native progress/revenue rows and filtered header totalRevenue; campaign_detail requires a campaign callout_id. Both include click/view and purchase/view/click percentages, preserving absent inputs as null. campaign_list supports complete=true; caps/errors fail clearly. Omitted campaign dates mean all period; explicit bounds use inclusive Unix ms. Forms page (/forms-ui/#/form-lists): use discovery, summary and detail. For a general web-widgets request, cover both forms and campaigns separately. Claspo reports use form widget_ids, variant_ids or integration_ids from their discovery/metadata. Their inclusive public date range is converted to the native exclusive end. Campaign and form identifiers are separate namespaces; do not mix their selectors. Default limit is 100 for campaigns, 50 for forms. Returns signed JSON; native nulls/currency remain intact and tokens stay private.
get_workflow_export
ChatGPTGet one workflow RawJson export for the authenticated Reteno organization. Use this tool when you need the original workflow definition as JSON. RawJson is stored as a graph array containing block nodes and Connection edges; generic Custom and CustomCondition blocks derive their meaning from taskName and taskParams. Fetches the current private campaign state and returns core workflow metadata, frontend-compatible startConfig and stopConfig fields, plus signed JSON download artifacts for the original RawJson graph and a separate configs file containing startConfig and stopConfig. The RawJson body is not returned inline. modelHash is the opaque precondition for update_workflow; retain the hash from the state you used to author the edit.
list_app_inbox_messages
ChatGPTSearch app inbox messages by partial name.
list_broadcasts
ChatGPTList campaigns. With private access, list scheduled campaigns using search, limit (1..100), offset, and embed. Use this read to reconcile an uncertain creation; multiple possible matches are not permission to create another campaign.
list_contacts
ChatGPTSearch contacts using the supplied query filters.
list_custom_blocks
ChatGPTlist_custom_blocks: find existing reusable email modules; read-only. This is the idempotency pre-check for module creation: the create endpoint's (category, name) does NOT upsert, so creating twice with an identical bundle mints two distinct modules. Call this first -- optionally with query to search by name/id -- and only create when nothing matching is already present. It also doubles as the post-hoc verification re-check for a create whose immediate readback came back verified=false: that miss is believed to be listing-index lag, and a later call here can find the module the create's own verify pass did not. uid is the customer-visible module ID; id is the numeric library ID used by place_email_module. Upstream query search may not match a UID. If it returns no match, broaden to module-name words or omit query, then match the supplied UID exactly against returned uid values. Use the matched numeric id internally. For missing or ambiguous matches, offer module names as choices; never ask users for numeric IDs. Read-only: no upload session, no write receipt. Returns {modules: [{id, uid, name, category, key, blockType, synchronizable, width, height}], total}. html/css/screenshotHtml are never included -- this tool exists to check existence, not to read module content. Capped at 200 entries; upstream category pages may return fewer. total includes reported rows outside those pages. If it exceeds the returned count, refine the name query before concluding a module is absent. Like module creation, the listing request uses the org's newest email message id for its identity header when one exists. For a fresh organization with no email messages, it uses numeric 0.
list_email_interfaces
ChatGPTGet existing public interfaces plus available, exactly formatted Reply-To choices. `replyToSelection is additive; the existing data` is unchanged. Stored values need not appear in these choices. "" means explicit clear.
list_email_messages
ChatGPTSearch regular emails or templates by partial name or tag. Query keys: `search (name fragment or tag), maxrows (1..500, default 500), and startindex (one-based, default 1). For example, query={"search": "Welcome", "maxrows": 500, "startindex": 1}. name, limit, and offset are unsupported. For the next page, increase startindex by maxrows. A short page ends the search; stop repeated identical pages. By default, returns regular emails only (template=false). Set template=true for Saved Templates only, or template=null` for both. Mixed results list regular emails first, then Saved Templates, with each group ordered by most recently updated. startindex/maxrows apply to that combined sequence, not separately to each group. Template and mixed searches use the account's content inventory and include totalCount; the public email endpoint does not include Saved Templates.
list_groups
ChatGPTList segments available in the account.
list_mobile_push_messages
ChatGPTSearch mobile push messages by partial name.
list_segment_definition_facets
ChatGPTList available dynamic-segment definition category values. The returned facets are limited to raw categoryType, groupType, group, conditionType, and type values from the same definition source as list_segment_definitions. segment_id and version only select the source/context; they are not facets.
list_segment_definitions
ChatGPTList live dynamic-segment condition definitions for the organization. Pass an existing dynamic segment ID when editing it so Reteno can return definitions in that segment's context. This calls the same Reteno definition source as eSputnik segment-condition's getDefinitionsView: optional version and groupId, returning ConditionDefinitionView items with column IDs, types, operators, options, support flags, and grouping metadata. Category filters are limited to raw categoryType, groupType, group, conditionType, and type. Each accepts one string or a list of strings and matches case-insensitively. query searches definition names, descriptions, labels, IDs, and option values. id exact-matches nested ID fields such as id, columnId, profileInputId, and option IDs. A match in object-shaped definition identity metadata returns the full definition. A match inside a parallel nested list, such as event options, returns the full parent definition with that list narrowed to unique matching entries; each retained entry remains complete. Responses using id also include idResolution with requested, matched, and unresolved IDs in request order. Definitions whose isSupported is not exactly true are always dropped after reading the backend definitions and before the other local filters. By default, returned items omit support flags and reduce operator objects to operator-name lists for compact analysis. Set raw=true to retain the full supported endpoint objects, including operator type/request metadata, support flags, options, and nested event-parameter definitions. Set for_sdk=true during segment authoring to stage the complete, unfiltered raw catalog and the authoritative eSNG UI context (timezone and resolved organization feature decisions) as a short-lived artifact while returning the compact view to the agent. The response then includes catalogRef; the trusted SDK runner downloads that exact snapshot without placing it in model context. Set segment_type=PARAMETRIC for a parametric authoring target. Like eSNG, the tool then reads the common definitions without a dynamic groupId and signs PARAMETRIC editor mode into the SDK catalog. Set include_details=true to attach resolved contact-field, message, tag, and event metadata implied by the final filtered definitions.
list_sms_interfaces
ChatGPTGet SMS interfaces.
list_sms_messages
ChatGPTSearch SMS messages by partial name or tag; returns up to 500 results per request.
list_workflows
ChatGPTList workflows available in the account.
patch_business_profile
ChatGPTPatch supported Business profile fields using a fresh read and full PUT. Objects merge recursively, arrays replace wholesale, and explicit empty strings or lists clear values. Leave an unknown language list out of the patch entirely; do NOT patch it to []. Divergent non-empty languages and brand.brandVoice.defaultLanguages are rejected after merging; change both together when changing an existing language pair. Never send a website extraction here: stage it with prepare_business_profile_upload and use save_business_profile, both with operation="extraction_replace". Extraction empties mean not found, whereas explicit patch empties clear values.
place_email_module
ChatGPTPlace a saved static library module in the email's HTML/CSS only. Composes from the saved public HTML/CSS using backend endpoints only. This method does not open a browser or update Stripo's native JSON model. After placement, the user must open the edited email in the browser editor to update the JSON model. For a translation, open that specific language variant. Before subsequent JSON editing, use get_email_model to read the refreshed model and hash; do not reuse a model downloaded before placement. With language, message_id is the base and the translation must exist. Select exactly one HTML td builder node (STRIPE/STRUCTURE/CONTAINER), using tag/class/HTML-id/quoted-attribute selectors and descendant or child combinators. Native JSON ids are not HTML ids. Before/after adds a row at a single-cell builder row; ambiguous layouts, dynamic product configurations and CSS affecting unrelated content are refused. Saved HTML/CSS is the sole composition and content-concurrency source; no populated native model or saved/native agreement is required. Existing saved CSS is retained. Only CSS used by the inserted subtree is imported; shared rules may be narrowed using existing module-local classes/IDs, including selector groups for an outer element and its descendants. Placements that cannot isolate the required styles are refused. Readback tolerates normal formatting outside whitespace-sensitive content and provably unused generated CSS, never changed content, active styles, metadata, or synchronization bindings. Identical requests are read-only replays, including after a restart. If an attempted write cannot be verified, no further content PUT is made. Success requires readback of the saved HTML/CSS to match the composed module content, library ID, synchronization markers, unrelated content/styles and preserved metadata. The HTML write receipt has verificationScope=saved_html_css and no modelHash. Native JSON may still be stale; it is neither read nor verified here. nextAction identifies the exact email/translation to open manually, followed by a fresh get_email_model check of its module content/binding before JSON editing. Use list_custom_blocks/get_email_model to discover modules and models.
prepare_business_profile_upload
ChatGPTPrepare one JSON-file upload for a Business profile save. Choose profile_update for a profile document, or extraction_replace for finalized website extraction (the compatible default). The operation is bound to the session and cannot be changed at save time. Returns {uploadSessionId, uploadUrl, operation, ...}; PUT the file by path as application/json, then save_business_profile with the same operation. Sessions are client-bound and expire after 1800 seconds by default. Ordinary OAuth preparation also binds the verified organization. Available to authenticated Reteno users; other brands require an agent token. Preparation and upload do not persist the profile.
prepare_email_message_upload
ChatGPTPrepare a short-lived upload session for email HTML and CSS.
prepare_email_model_upload
ChatGPTPrepare a short-lived upload session for one structured model file. Returns an upload session (uploadSessionId + a signed upload URL) for the SDK-edited {settings, stripes} model FILE. Upload the file via the proxy (PUT, application/json), then finalize with update_email_model by reference. operation="update" targets an existing message; operation="create" is a from-scratch compose and is accepted ONLY for a registry-bound shell minted via create_email_shell (owned by this client, unexpired). The model is NEVER a tool argument. The uploaded complete model is authoritative; no browser action plan is accepted or replayed. For concurrent edits, prepare two operation="update" sessions: upload the edited target to one and the unmodified get_email_model file to the other. Pass their IDs as upload_session_id and base_upload_session_id to update_email_model. Each session holds one complete JSON document.
prepare_image_upload
ChatGPTPrepare a short-lived upload session for one image file.
prepare_workflow_upload
ChatGPTPrepare a short-lived upload session for workflow graph JSON and optional configs JSON. Available to authenticated Reteno, Yespo, and eSputnik users. No agent token or organization allowlist entry is required.
resume_email_sender_update
ChatGPTResume only the sender step of this caller's previously created email. Use the recovery arguments returned by create_email_message. The consumed upload's ID resolves server-owned organization/message/sender intent; callers cannot choose a different target or sender. No upload, public content PUT, or create POST is repeated. Completed retries read back without undoing later edits. Available to ordinary MCP clients; arbitrary private metadata updates remain agent-only.
save_business_profile
ChatGPTSave one uploaded JSON file with its explicitly prepared operation. profile_update requires a profile document with a brand object and no website argument. It reads current storage, recursively merges supplied object fields before validation defaults, and PUTs. Omitted fields and open extensions in the normalized stored profile survive. Stored importantLinks rows lacking a non-empty name or usable URL are dropped by read normalization and are not preserved by omission. Arrays replace wholesale; supplied values can overwrite or clear. Null is not omission: blankable strings become empty strings, some string lists become [], and typed collections such as brand.logos reject null. Omit unknown language lists; do NOT send []. Divergent non-empty languages and brand.brandVoice.defaultLanguages are rejected after merging; change both together when changing an existing language pair. extraction_replace requires website matching the staged extraction's host; it normalizes extraction over an empty base and POSTs, clearing omitted saved values. It returns website, cleared, dropped and sanitized notices. First prepare_business_profile_upload with the same operation and PUT the JSON file to uploadUrl. Success consumes the session; failures do not. profile_update returns orgId and brandkit, plus applicable warnings. Upload as application/json. Both operations accept ordinary Reteno OAuth; Yespo/eSputnik require the agent token. The host's write approval applies.
send_broadcast
ChatGPTStart a broadcast for email, SMS, mobile push, or web push. groups is the audience: plain group ids, at least one. A segment is addressed by its group id like any other group. The audience is required and cannot be empty -- the API rejects an empty one, and on some deployments reports that rejection as a bodyless 500 that reads like an outage, so this signature refuses it here where the reason is legible. start_date ("YYYY-MM-DDTHH:mm") schedules it; without it the broadcast starts immediately. batch_size with batch_interval_unit throttles delivery. domain applies to web push, app_id to mobile push and Telegram. extra is merged into the request body for any field this signature does not name yet; named arguments win over the same key in extra. For callers with private email campaign access, use_contact_timezone schedules start_date in each contact's local time (missing contact zones fall back to the account zone). It requires a date and forbids batching. Otherwise start_date uses the account timezone. The private path verifies saved settings, resolves audience IDs internally, and rejects unmapped extra fields. WEEK is available there; MONTH remains on the legacy path. A creation error may carry an existing broadcast ID or an uncertain outcome: read/reconcile it before retrying, never blindly create again. Public contract: docs.esputnik.com/reference/createbroadcast-1.
send_email_message
ChatGPTSend an email to a contact; if the email address does not exist yet, Reteno creates a new contact. Required payload fields: from (one of the org's valid sender addresses), subject, htmlText, and emails (a non-empty array of recipient addresses). Optional: plainText, ampHtmlText, tags, externalRequestId, skipPersonalisation, subscriptionsKeys, trackUrls (default true). Contract read from docs.esputnik.com/reference/sendemail-1.
send_sms_message
ChatGPTSend an SMS to a contact; if the phone number does not exist yet, Reteno creates a new contact. Required payload fields: from (one of the org's valid sender names), text, and EITHER phoneNumbers (array) OR groupId (integer) -- supply exactly one of those two. Optional: tags, externalRequestId, subscriptionsKeys, trackUrls (default false). Contract read from docs.esputnik.com/reference/sendsms-1.
smart_send_message
ChatGPTSend a prepared message to one or many contacts, with optional per-contact personalization. Each recipient identifies ONE contact by exactly one of contactId, externalCustomerId, or locator (an email, phone, token, or chat id), and may carry jsonParam, externalRequestId, and languageCode. email selects the media channel (true for email, false for the rest); from_name overrides the sender identification. extra is merged into the request body for any field this signature does not name yet; named arguments win over the same key in extra. Contract read from docs.esputnik.com/reference/sendextendedpreparedmessage-1.
subscribe_contact
ChatGPTCreate or update a contact from a subscription form; new contacts are created as unconfirmed.
update_app_inbox_message
ChatGPTUpdate a base app inbox message.
update_app_inbox_message_translation
ChatGPTUpdate a translated app inbox message variant.
update_brandkit
ChatGPTWrite the saved Brand Kit for the authenticated organization. Send a COMPLETE Brand Kit document after reading it with get_brandkit. PUT preserves the stored value of any field the body omits; POST instead resets every omitted field to its default, but validation supplies defaults for omitted modeled fields before PUT and may therefore clear saved values. Keys the endpoint does not model are PERSISTED. If a language list is unknown, OMIT the key; do NOT send []. This is the temporary legacy alias of update_business_profile.
update_brandkit_patch
ChatGPTPatch supported sections of the saved Brand Kit. Unmentioned sections survive; arrays replace wholesale and explicit empty values clear fields. If a language list is unknown, leave it out of the patch entirely; do NOT patch it to []. This is the temporary legacy alias of patch_business_profile.
update_business_profile
ChatGPTWrite a complete Business profile document using existing PUT semantics. Read first with get_business_profile and preserve unrelated fields; use patch_business_profile for supported field changes. Explicitly empty lists clear those fields. Omitted modeled fields receive validation defaults and may therefore clear saved values. Omit an unknown language list; do NOT send []. Divergent non-empty languages and brand.brandVoice.defaultLanguages are rejected; change both together when changing an existing language pair. Never send a website extraction here: stage its JSON file with prepare_business_profile_upload and use save_business_profile, both with operation="extraction_replace", which applies extraction semantics and reports cleared fields.
update_contact
ChatGPTUpdate a contact by ID; profile changes may trigger configured workflows.
update_dynamic_segment
ChatGPTSave Segment Designer SDK UI-mode JSON to an existing dynamic segment. Like the eSNG editor, this sends the new schema to the dynamic group's schema PUT endpoint, then fetches the segment to confirm that it remains retrievable. The backend-normalized schema is not compared with the submitted UI schema. Authenticated Reteno, Yespo, and eSputnik users can see and call this tool without an agent token or organization allowlist entry.
update_email_message
ChatGPTUpdate an existing regular email or template by internal message ID. A payload containing only name renames the entire language family using its base message ID; variant IDs are rejected. The rename is verified without an upload session or HTML/CSS export. Content updates require payload.uploadSessionId from prepare_email_message_upload. The resource classification is preserved because `template` is never sent by this update tool; classification is selected only during create.
update_email_message_translation
ChatGPTLocalize content while preserving the target's Reply-To and sender. Omit payload.replyTo to preserve an existing locale (including blank), or inherit the base for a new locale. A string sets exactly that value; "" explicitly clears; null is invalid. Reply-To uses a private metadata PATCH to a verified actual variant ID, never the public translation PUT. Omit payload.sender/from to preserve the sender (inherit for a new locale). payload.sender sets an exact private value; dynamic from expressions use the same private path. If the public content API needs a registered interface, it is temporary and the intended sender is restored on the actual variant. Sharing flags are never changed. Prepared language-bound sessions and legacy inline HTML are supported. Pending intent survives new upload sessions and server restarts. Retry identical content/metadata; optional payload.replyToOperationId gives a stable logical key across transport retries (reuse it with identical input). A conflicting operation is refused while recovery is pending. Report completion only when persisted and verification.verified are true. Set top-level template=true and omit payload.from to use the upstream sender-optional localization mode, including with prepared uploads. The flag uses upstream template semantics for the variant; it is never enabled automatically. Static from values are still validated upstream. A deleted variant's old operation cannot be replayed; recreating it needs a fresh upload and a new replyToOperationId.
update_email_metadata
ChatGPTRename an email or update subject/preheader, Reply-To, or sender. name requires the base message ID and renames the entire language family through the private common-fields PATCH; variant IDs are rejected with base ID guidance. Fresh read-back verifies the exact name on every family member, a fresh base revision, and all other fields unchanged except write timestamps. No HTML/CSS upload, export, public PUT, or model write is used. Omission preserves the name; blank/null are invalid. Send name separately from other metadata fields. A retry skips an already-applied rename. A subject update uses the organization-scoped private content PATCH /content/rest/v3/message/email/{actualMessageId}/content with subjects only. It replaces the first nonblank subject (or initializes an empty list), retaining the other variants. Success requires fresh private readback of subjects and unchanged editable rawHtml/CSS, other metadata and locales. No public PUT, body compilation, or model write is used; there is no public fallback. A retry reads first and skips an already applied subject PATCH. Failed or unavailable verification is an error, even after a successful PATCH acknowledgement. reply_to is optional: omission preserves, a string sets the exact value, "" clears, and null is invalid. Address a locale by its actual messageId from get_email_message_export. Reply-To uses the private metadata PATCH and fresh read-back, with no HTML/CSS upload. Subject plus sender/Reply-To uses separate content and common-fields PATCHes and can partially succeed. Subject plus preheader is rejected before any write: preheader is embedded in HTML and has no verified content-field PATCH contract. Preheader-only calls retain the legacy public PUT acknowledgement, without a body-preservation or preheader-persistence guarantee. For native-model emails, edit model.metadata.preheader and persist with update_email_model, then verify a fresh get_email_model download and the saved hidden preheader in get_email_message_export. sender accepts a nonempty literal string (including dynamic expressions); omission preserves it, blank/null are invalid. Sender uses the same private PATCH and readback. No sharing flags are changed. Saving does not verify dynamic sender feature enablement, verified domains, or delivery readiness.
update_email_model
ChatGPTFinalize an uploaded structured email model without opening a browser. To update the downloaded native JSON model, first call `prepare_email_model_upload, upload the JSON file to its signed URL, then pass the returned upload_session_id here. The model is never an inline tool argument. The staged model passes the compatibility, fidelity, hashing, idempotency, and persistence gates before any destructive write. The complete uploaded model is sent to Stripo's document-state API, converted to editable HTML/CSS by the Plugin Backend, and persisted through Reteno's Public API. Set native preheaders in model.metadata.preheader (metadata.preheader in the downloaded JSON). After saving, verify the preheader in a fresh get_email_model download and the persisted hidden preheader through get_email_message_export. Subject, Reply-To, and sender changes use update_email_metadata; its legacy flat preheader argument does not guarantee persistence and is not the native preheader editing path. For edits to an acquired model, upload its UNMODIFIED get_email_model file in a second operation="update" session and supply its ID as base_upload_session_id. Stripo merges only the base-to-target delta into the current document, preserving unrelated concurrent edits. expected_model_hash then verifies the uploaded base, not the live document. Without a base, the complete target replaces the live state and expected_model_hash retains its live-state concurrency check. A base is not supported for operation="create" or the legacy browser transport. Never swap the two uploads or drop the base after an error. Both sessions must belong to this client and message and are consumed together on a successful persist; dry runs and errors retain them. Merge acknowledgements cannot prove target equality: appliedModelHash is null and readbackRequired is true after a merged SET. Reacquire with get_email_model, verify the requested changes and preview, then use that fresh file as the next base. After an uncertain write, read first; never retry blindly or replace the document to resolve a merge error. Stripo styling notes: paragraph colour must be pinned with each text block's settings.fontColor (inline <p style=color> and inherited td colour can be overridden); tinted text panels need that block's settings.backgroundColor. Image blocks must not carry settings.backgroundColor (use the container), and small logos/icons should set responsiveMobile:false to avoid full-width mobile output. A line-mode spacer with height is rejected locally. Columns with multiple stacked containers are supported by document-state updates and validated against Stripo's deployed schema. The single-container limit applies only to the legacy browser import. Concurrent-retry contract (create mode): a create finalize CLAIMS its registry shell atomically. If a DUPLICATE finalize for the same shell arrives while the first is still in flight, it is refused with a typed shell_in_flight` error (retryable) naming the idempotency key -- retry the SAME finalize once the first completes; do NOT re-mint via create_email_shell (that composes a duplicate email). A duplicate carrying the same idempotency_key that arrives AFTER the first finalize has completed replays the stored success (no second write).
update_mobile_push_message
ChatGPTUpdate a base mobile push message through the public replacement API. Read the message first. payload must include name, title and text, even when only one of them changes. Preserve the acquired name and all unrequested fields. The public fields include link, imageSource, ttlSec and customData; customData is a JSON-encoded string, not an object. Merge requested Custom Data keys into that decoded string and retain all other keys. Do not copy the GET-only languages or translations fields into payload. Update other language variants separately. A malformed payload is not a permission denial.
update_mobile_push_message_translation
ChatGPTUpdate a translated mobile push message variant. Read the current variant first. The public payload requires name, title and text; retain its name and all unrequested fields. Use link, imageSource, ttlSec and JSON-string customData from the public message representation. Preserve other locales and other Custom Data keys. Do not include the GET-only languages or translations fields in payload. A malformed payload is not a permission denial.
update_mobile_push_token_activation
ChatGPTUpdate mobile push token activation state.
update_sms_message
ChatGPTUpdate a base SMS message.
update_sms_message_translation
ChatGPTUpdate a translated SMS message variant.
update_workflow
ChatGPTUpdate a workflow using SDK code or an uploaded graph, with a stale-state guard. First call get_workflow_export and retain modelHash as expected_model_hash. Provide exactly one of code or uploadSessionId. SDK code runs in the internal workflow browser service against the current graph/start/stop configuration; finish scripts with sdk.finish(). Existing uploaded graph/config files use the same guarded persistence and read-back path. No activation is performed. For a new workflow, call create_workflow once, retain its workflowId, export that paused shell and then update it. On workflow_conflict, re-read and reconcile the intended edit. Do not substitute a fresh hash and blindly replay code. commitStatus=unknown means a write may have succeeded: inspect live state before any further mutation. No idempotency key is required. workflow_editor_at_capacity means this call did not execute or save. The editor admits one waiting request for up to ten seconds. Overflow or admission expiry returns workflow_editor_at_capacity without execution. Wait error.retryAfterSeconds, then retry the same arguments including expected_model_hash. MCP does not automatically retry capacity failures. A classified editor bootstrap timeout gets one internal transform-only retry. Editor failures return failurePhase=editor, commitStatus=not_committed; this applies to this update only, never to an earlier create. Exhausted bootstrap retries retain the existing workflow ID; do not recreate it.
upload_contacts
ChatGPTImport contacts from an external file URL; optional events can trigger workflows.
upload_image
ChatGPTUpload a previously staged image file to Reteno file storage. If the staged file is an SVG (which email clients cannot render), it is rasterized to PNG server-side and hosted with a `.png URL. purpose: 'generic' rasterizes only (still refuses uncovered-glyph text so tofu never ships); 'logo' additionally applies an aggressive fail-closed quality gate. On conversion failure the tool RETURNS {"error_code": ..., "reason": ...}` (never raises, never uploads the SVG, never substitutes a favicon) so the caller can defer autonomously.
upsert_contact
ChatGPTCreate or update one contact synchronously by its identifiers; field changes can trigger workflows. A successful response means the profile is already updated.