MCP App Store
by KeenEthics
Realize Plugin icon

Realize Plugin MCP Server for Claude and ChatGPT

By View source

Overview

Query Realize campaigns, pull performance reports, and create/update campaigns, ad items, and conversion rules through natural language — powered by the Realize remote MCP.

How to connect

  1. Claude: Open the Realize Plugin plugin listing for Claude and install it in Claude Code. Claude listing
  2. ChatGPT: Open the Realize Plugin listing in the ChatGPT app directory and choose Connect. ChatGPT listing Realize Plugin exposes 33 tools here.

Tools

create_audience

ChatGPT
Create an audience on a Realize account. audience_type picks the kind and decides which fields apply: PIXEL — a rule over Taboola Pixel traffic. Requires display_name, type, category, event_name, status, condition, effects, include_in_total_conversions and include_in_total_value. Send category NONE, status ACTIVE, both include_* false and effects [] unless you have a reason otherwise — that is what the console sends. A BASIC audience with no URL condition takes condition {}. type, category and event_name are IMMUTABLE after creation, so get them right here; everything else can be changed later with update_audience. COMBINED — a boolean combination of audiences that already exist. Requires audience_name and combined_audience_components (at most 4 components, at most one NOT). Member ids come from the ordinary search_audiences listing; most of them work. One is rejected with "Udd with id N doesn't exist" when that audience has no underlying segment yet, typical of one created moments earlier. search_audiences(combinable=true) then lists ids known to work — a different and largely third-party population, so treat it as the fallback rather than the first stop. The server sets country and platforms on a combined audience itself; sending them is not required. CUSTOMER_LIST — built from a CSV you supply. Requires audience_name, crm_targeting_type, crm_file_type and file_url (an https link to the file; the server downloads it and uploads it to Realize). Give the link rather than pasting rows: the contents never pass through the conversation. The audience targets or suppresses the people in the file. Matching takes time, so the new audience will be small or empty right after creation. FILE FORMAT — one identifier per line, one column, no header row, encoded as UTF-8; a file in another encoding is refused rather than uploaded unread. Minimum 1,000 records (the console's guidelines recommend 2,000+ for scale) and at most 3,000,000 — though the download size limit usually binds first, at roughly 800,000 hashed emails or 1.4 million device IDs, so only short identifiers such as ZIP codes reach the record ceiling. Smaller files are still forwarded, and the result says so when one is under 1,000. EMAIL plain addresses are hashed to SHA256 for you before upload, so send the file you have; already-SHA256 files pass through, and MD5 or SHA1 are rejected by Realize. MOBILE_ADVERTISING_ID not hashed — send the raw IDs. ZIP_CODE not hashed — 5-digit US ZIP codes. Names must be unique on the account and 3-100 characters. Unknown field names are rejected rather than ignored, and so are fields belonging to another audience_type, so a mistake fails loudly instead of creating an audience missing that value. Returns the created audience including its server-assigned id. DO NOT RELY ON THE RETURNED ID BEING READABLE. A new audience is often not readable by id at all — observed unreadable for at least 12 minutes, and not known to recover on its own — while appearing in the plain listing straight away. A 404 therefore does NOT mean the create failed and does NOT mean you chose the wrong audience_type; confirm from the listing instead, and do not recreate. For CUSTOMER_LIST the returned id is the underlying rule's, which is a different id from the one the listing shows and is not readable at all — find it in the listing by name.

create_campaign

ChatGPT
Create a campaign on a Realize account, with full targeting in one call. Returns the created campaign (id, status=PAUSED); ships paused, won't serve until items are added (set is_active=true to launch). Prerequisite: call search_accounts first to obtain a valid account_id. The numeric account ID is rejected. Targeting blocks are FULL-REPLACE: sending audiences_targeting with one audience wipes the others; sending country_targeting replaces all country targeting. To edit a single entry, read with get_campaign, modify locally, send the merged result. Scalar fields (name, cpc, is_active, …) are partial-merge — omitted scalars keep their prior value. Geo: send any subset of country_targeting, region_country_targeting, dma_country_targeting, city_targeting, postal_code_targeting. Sub-dimension mutex: at most one of region/dma/city/postal_code per campaign; INCLUDE country must already be set before a sub-dimension is accepted. All amounts (spending_limit, daily_cap, cpc, cpa_goal, cpc_cap) are numbers in the account's default currency. Conditional rules: - spending_limit_model is required, EXCEPT with campaign_group_id: the campaign then takes the group's budget when you omit it. In an AUTOMATIC group you MUST omit spending_limit_model, spending_limit, daily_cap and daily_ad_delivery_model — the group owns them and the API rejects them. - spending_limit_model = MONTHLY|ENTIRE → also send spending_limit. NONE → also send daily_cap. - daily_ad_delivery_model = STRICT → also send daily_cap. BALANCED → omit daily_cap. - bid_strategy = MAX_CONVERSIONS → cpc_cap allowed. Other strategies → omit cpc_cap. - bid_strategy = TARGET_CPA|MAX_CONVERSIONS|MAX_VALUE → omit publisher_bid_modifier. - marketing_objective = BRAND_AWARENESS|DRIVE_WEBSITE_TRAFFIC → send cpc; bid_strategy SMART (default) or FIXED; omit cpa_goal. - marketing_objective = LEADS_GENERATION|ONLINE_PURCHASES|MOBILE_APP_INSTALL → bid_strategy = TARGET_CPA|MAX_CONVERSIONS|MAX_VALUE; if TARGET_CPA also send cpa_goal; omit cpc. - pricing_model = VCPM → bid_strategy MUST be FIXED; cpc carries the vCPM rate; omit cpc_cap and cpa_goal; traffic_allocation_mode is forced to OPTIMIZED server-side. - pricing_model omitted → server defaults to CPC. - If both start_date and end_date sent: end_date >= start_date. Discovery (call before constructing the payload to resolve IDs/names): - search_geos → country_targeting / region_country_targeting / dma_country_targeting / city_targeting / postal_code_targeting values. Each result is {code, name}; use code (e.g. "CA"), not name ("California"). - search_techno → os_targeting sub_categories and browser_targeting values. (platform / connection_type / os_family enums are listed inline in their schema descriptions.) - search_audiences → audiences_targeting audience IDs. - search_lookalike_audiences → lookalike_audience_targeting rule_ids. - search_contextual_segments → contextual_segments_targeting segment IDs. - search_publishers → publisher_targeting and publisher_bid_modifier.target values. Each result is {id, name, account_id, ...}; use account_id (e.g. "linkvertise-channel"), not name ("Linkvertise - Channel"). - get_conversion_rules → conversion_rules.rules IDs. - list_time_zones → activity_schedule.time_zone IANA names. TIMEZONE — user's OWN / "my" / "local" / "user" / "viewer" time: set activity_schedule.time_zone to the literal string "USER_TIMEZONE". Do NOT run a shell command, read the host clock, call list_time_zones, or resolve it to an IANA name. Only use an IANA name when the user names a specific place (e.g. "New York time"). CORRECT: "Saturday 9am-9pm in my local timezone" -> "time_zone": "USER_TIMEZONE" INCORRECT: (detect host TZ, then) -> "time_zone": "America/New_York" Read-only — NEVER send: id, advertiser_id, status, approval_state, spent, policy_review, target_cpa, target_cpa_learning_status. (target_cpa is server-recommended; user goal is cpa_goal.) All targeting (including audiences, lookalike, contextual segments) goes in one atomic call; either the whole campaign with al…

create_campaign_group

ChatGPT
Create a campaign group on a Realize account. A group holds a set of campaigns: it sets their shared goal (objective, KPI, conversion rules), flight dates and — optionally — one budget shared across them. Put campaigns in it with create_campaign / update_campaign (campaign_group_id). Ask the user, in this order, as the Realize UI does: 1. marketing_objective, then a kpi the objective allows: BRAND_AWARENESS: CTR, CPC, CPM DRIVE_WEBSITE_TRAFFIC: CTR, CPC, CPQV LEADS_GENERATION: CTR, CPC, CPA, CPL, ROAS ONLINE_PURCHASES: ROAS, CPA, CTR, CPC MOBILE_APP_INSTALL: CPI, ROAS, CPA, CTR, CPC 2. conversions: the account default, or rules picked manually (conversion_rules). 3. optional CPA goals for recommendations (kpi_data) — mind its field names. 4. dates and budget (budget_distribution_method, spending_limit_model + amounts). Required: name and spending_limit_model (even with a CUSTOM budget split). marketing_objective and kpi go together: send both or neither. spending_limit_model NONE needs a daily_cap. The response notes any change the tool made to the request. Fields the API manages itself (status, bid_strategy, cpa_goal, language) cannot be set. Returns the created group with its id.

create_conversion_rule

ChatGPT
Create a conversion rule on a Realize account. The rule is always created ACTIVE, as in the Realize UI — there is no status parameter. display_name must be unique per account. An event can hold only one ACTIVE rule, so call get_conversion_rules first; if the event already has one, update that rule with update_conversion_rule instead of creating another. type/category/event_name are immutable after creation. Retire a rule later via update_conversion_rule with status DISABLED/ARCHIVED (there is no delete). Unknown field names are rejected rather than ignored, so a misspelling fails loudly instead of creating a rule missing that value. Returns the created rule, including its server-assigned id — use that id for get_conversion_rules, update_conversion_rule, or conversion_rules.rules on a campaign.

create_display_item

ChatGPT
Create a display item directly attached to a campaign on Realize. "Item" and "ad" refer to the same object — this tool creates one. Returns the created item with its server-assigned id, status, and approval_state. Prerequisites: call list_campaigns or get_campaign to obtain campaign_id and the owning account_id (the campaign's advertiser_id — a NETWORK/parent account is rejected with 403 here even though it can list the campaign). Numeric account IDs are rejected. Two creative modes — pass exactly one discriminator: - Third-party (3P) tag: send ad_tag (raw HTML/JS) + dimensions. Realize validates structure server-side and performs ${CLICK_URL} macro replacement on submit. Soft 16 KiB cap client-side. - First-party (1P) Realize-hosted asset: send asset_url (public https URL). Realize ingests by URL file extension — image (jpg/jpeg/png/gif/bmp/webp), video (mp4/mpg/mpeg/flv/webm), or HTML5 bundle (zip; origin must serve Content-Type: application/zip). Do NOT send dimensions with asset_url; Realize populates them from the asset. Max 200 MB. For URL-crawled native items use create_native_item instead. HOSTED_SOCIAL is not currently supported. Not applicable to display creatives: branding_text and cta — the ad tag or hosted asset handles its own branding and click; sending these has no effect. Read-only — NEVER send: id, campaign_id (set via path), creative_type (server enforces DISPLAY), display_data.display_ad_type (server-detected), status, approval_state, learning_state, policy_review. Item creation goes in one atomic call; either the item commits with all fields, or it doesn't. Comprehensive examples (only account_id, campaign_id, url, creative_name, and exactly one of ad_tag or asset_url are mandatory). New items are always created active; use update_display_item to pause if needed. Example — 3P tag (300x250 CM360 banner): { "account_id": "acme-inc", "campaign_id": "49184816", "url": "https://example.com/landing", "ad_tag": "<ins class=\"dcmads\" style=\"display:inline-block;width:300px;height:250px\" data-dcm-placement=\"N12345.1234567ACME/B12345678.123456789\" data-dcm-rendering-mode=\"script\" data-dcm-https-only data-dcm-resettable-device-id=\"\" data-dcm-app-id=\"\" data-dcm-click-tracker=\"${CLICK_URL}\"><script src=\"https://www.googletagservices.com/dcm/dcmads.js\"></script></ins>", "dimensions": [{"width": 300, "height": 250}], "creative_name": "Acme Spring Sale 300x250" } Example — 1P hosted image (image type chosen from .png extension): { "account_id": "acme-inc", "campaign_id": "49184816", "url": "https://example.com/landing", "asset_url": "https://cdn.example.com/creatives/acme-spring-300x250.png", "creative_name": "Acme Spring Sale 300x250 — hosted image" } Example — 1P HTML5 bundle (.zip URL; origin must serve Content-Type: application/zip): { "account_id": "acme-inc", "campaign_id": "49184816", "url": "https://example.com/landing", "asset_url": "https://cdn.example.com/creatives/acme-spring-300x250.zip", "creative_name": "Acme Spring Sale 300x250 — html5" }

create_native_item

ChatGPT
Create a native item directly attached to a campaign on Realize. "Item" and "ad" refer to the same object — this tool creates one. Returns the created item with its server-assigned id, status, and approval_state. Prerequisites: call list_campaigns or get_campaign to obtain campaign_id and the owning account_id (the campaign's advertiser_id — a NETWORK/parent account is rejected with 403 here even though it can list the campaign). Numeric account IDs are rejected. Scope: this tool creates a native image ad directly attached to the campaign, a carousel (send cards), or a performance video — a native video ad (send video_url). For third-party display ads (banner ad tag with fixed dimensions) or first-party Realize-hosted display assets (image/video/HTML5 zip URL with fixed dimensions), use create_display_item. RSS feed items and the Creative Library are not supported. Discovery (call before constructing the payload to resolve IDs/names): - list_cta_types → cta.cta_type allowed values. Read-only — NEVER send: id, campaign_id (set via path), type, status, approval_state, learning_state, orientation, policy_review. Item creation goes in one atomic call; either the item commits with all fields, or it doesn't. Required fields: account_id, campaign_id, url, title, description, thumbnail_url. Optional: branding_text, creative_name, cta. New items typically transition PENDING_APPROVAL → RUNNING. Carousel: send cards (3-10 cards, each with url, title, description, thumbnail_url and optional cta / creative_name) to create one carousel ad whose cards rotate. The carousel's own url / title / thumbnail_url are then optional and default to the first card's; description and cta belong on the cards. Optional fallback_policy. A carousel needs a campaign with no other items. Call list_items first, and if the campaign already has items, create a new campaign. Carousel cards cannot be scheduled. Performance video: send video_url (a publicly reachable direct link to the video file). url and title are required; description, thumbnail_url (fallback image — generated from the video if omitted), branding_text, creative_name and cta are optional. Backstage checks the link and copies the video during the call. A link it cannot use fails with a 400 — "URL is not accessible" (missing file, or a page that is not a video) or "Invalid video_url" (host does not resolve); fix the link, don't retry. A 500 whose message starts "Failed to upload video" or "no fallback image for video" means the file was fetched but could not be processed — try a different file rather than retrying. Cannot be combined with cards. Plain English: a "Spring Sale" ad for the Acme brand, landing at example.com/landing, with explicit headline / body / thumbnail and Shop Now CTA. New items are always created active; use update_native_item to pause if needed. { "account_id": "acme-inc", "campaign_id": "49184816", "url": "https://example.com/landing", "title": "Save 20% This Spring", "description": "Limited-time offer on all spring collection items.", "thumbnail_url": "https://cdn.example.com/spring.jpg", "branding_text": "Acme", "creative_name": "Acme Spring Sale Native", "cta": {"cta_type": "SHOP_NOW"} } Example — carousel: { "account_id": "acme-inc", "campaign_id": "49184816", "creative_name": "Acme Spring Carousel", "fallback_policy": "FIRST_CHILD", "cards": [ {"url": "https://example.com/p/1", "title": "Linen shirt", "description": "Breathable, 20% off.", "thumbnail_url": "https://cdn.example.com/1.jpg", "cta": {"cta_type": "SHOP_NOW"}}, {"url": "https://example.com/p/2", "title": "Canvas bag", "description": "Carry everything.", "thumbnail_url": "https://cdn.example.com/2.jpg"}, {"url": "https://example.com/p/3", "title": "Straw hat", "description": "Made for summer.", "thumbnail_url": "https://cdn.example.com/3.jpg"} ] } Example — performance video: { "account_id": "acme-inc", "campaign_id": "49184816", "url": "https://example.com/landing", "title": "See the spring collection in motion", "video_url": "https://cdn.example.c…

get_campaign

ChatGPT
Get details for one campaign on a Realize account. CONVERSION GOAL: a campaign's conversion_rules lists only rules set on the campaign itself. The conversion goal is usually set on the campaign group instead, so an empty conversion_rules does NOT mean the campaign has no goal. When campaign_group_id is set and conversion_rules.rules is empty, call get_campaign_groups with campaign_group_id set to that id — the group's conversion_rules are the goal this campaign optimizes toward. Only when neither has rules does the campaign fall back to the account default (the account's active rules marked include_in_total_conversions, see get_conversion_rules). Rules set on the campaign replace the group's for that campaign; they are not added to them.

get_campaign_breakdown_report

ChatGPT
Get the campaign breakdown report for an account. One row per campaign (id column campaign). Returns CSV with per-campaign metrics. One call per page returns complete data. Works for ALL account types, including GROUP and admin-network accounts that the dynamic report (get_dynamic_report_data) cannot serve — use this tool for those. Check Total in the summary line for the full record count across all pages. Each row is one pre-aggregated record at the grain named on the Row key: line — the set of columns that together make a row unique. Identify columns by name, not by position (column order is not guaranteed). The metrics ctr, cpc, cpm, cpa, cvr and roas are computed server-side per row: use each value as-is — never recompute or average them across rows. To aggregate volume, sum only the raw counters (clicks, impressions, spent). For an overall total or rate, rely on the Total in the summary line (authoritative) — never sum rows across pages. If this call times out, retry it with the same arguments: the report continues generating and is cached upstream, so a repeat call usually returns it quickly.

get_campaign_groups

ChatGPT
Read the campaign groups on a Realize account (read-only). A campaign group holds a set of campaigns and is where their conversion goal is usually set: its conversion_rules apply to every campaign in it that has no conversion rules of its own. Each group also carries its budget (which may be shared by its campaigns), flight dates, status and KPI target. Set campaign_group_id to read one group — take it from a campaign's campaign_group_id (get_campaign / list_campaigns). Omit it to list groups, e.g. to find a group's id by name: one page per call, page / page_size (defaults 1/20, max 50), ordered by id; the response states page, page_size and the true total, so check total before assuming you have seen every group. Narrow with statuses or search_text instead of paging where you can. Both forms return the same {account_id, ..., values: [...]} shape.

get_campaign_history_report

ChatGPT
Get the campaign change history for an account. A change/audit log, NOT a performance report: one row per change event (campaign_id, change_time, id) — columns change_type, old_value, new_value, performer. No impression/click/spend or rate metrics. One call per page returns complete data. Check Total in the summary line for the full record count across all pages. If this call times out, retry it with the same arguments: the report continues generating and is cached upstream, so a repeat call usually returns it quickly.

get_campaign_reach_estimate

ChatGPT
Estimate the potential reach — audience size, how many users or impressions — of a hypothetical campaign before launch on a Realize account. Answers "how big is my audience / how many people will this campaign reach" for a given targeting + bid + budget. Forecast from the supplied config, not from historical performance. Reach narrows as more inputs are supplied: - Targeting only → audience reach - + cpc (with pricing_model=CPC, the default) → narrowed by bid competitiveness - + spending_limit or daily_cap → additionally capped by impressions the budget can afford cpc only affects reach when pricing_model=CPC. Budgets only affect reach when paired with cpc. Other combinations return audience reach. Returns lower/upper bound for each requested estimation type (IMPRESSIONS, MONTHLY_USERS).

get_conversion_rules

ChatGPT
Read the conversion rules on a Realize account (read-only). Omit rule_id to list the account's conversion rules (the backend returns ACTION rules only — pixel audience rules are not included here); the IDs returned populate conversion_rules.rules: [{id}] on create_campaign / update_campaign (LEADS_GENERATION and ONLINE_PURCHASES campaigns typically require at least one). The listing is paginated — one page per call, page / page_size (defaults 1/25, max 50), ordered by rule id so pages are stable — and each rule comes back in full with its condition and effects. The response states page, page_size and the true total, so check total before assuming you have seen every rule. Narrow instead of paging where you can: status filters to ACTIVE/DISABLED/ARCHIVED, search_text matches display_name, and rule_id returns one known rule (paging does not apply then). Both forms return the same {account_id, ..., values: [...]} shape; rule_id just yields a one-entry list. On a NETWORK/parent account the response also covers child accounts' rules — each rule's advertiser_id names its owner. Read a rule before updating it to see its current values, then send update_conversion_rule only the fields you want to change — the read payload itself is not a valid update body.

get_dynamic_report_data

ChatGPT
Execute a dynamic report query and return results as CSV (read-only) with a summary banner (Records, row Grain, and pagination), matching the other Realize report tools. ACCOUNT TYPES: serves PARTNER and NETWORK accounts only; for a GROUP or admin-network account this returns 403 — use get_campaign_breakdown_report (a classic report tool) for those instead. IMPORTANT: You MUST call get_dynamic_report_settings FIRST and use the exact fully qualified names from the metamodel - do NOT guess or fabricate field names. WORKING EXAMPLE: {"account_id": "<id>", "columns": ["PERFORMANCE_REPORT.CAMPAIGN.CAMPAIGN_NAME", "PERFORMANCE_REPORT.METRICS.CLICKS", "PERFORMANCE_REPORT.METRICS.SPENT"], "date_preset": "LAST_7_DAYS"} DATE RANGE (required): pass EITHER date_preset (e.g. LAST_7_DAYS) OR a custom range date_from + date_to in yyyy-MM-dd (e.g. "2026-01-01" to "2026-01-31") - not both. A custom range may span at most 12 months; for the last twelve months use date_preset=LAST_12_MONTHS instead. WORKFLOW: 1) search_accounts 2) get_dynamic_report_settings to discover valid field names 3) Pass those exact fully qualified names as columns here.

get_dynamic_report_settings

ChatGPT
REQUIRED FIRST STEP for dynamic reports (read-only). PARTNER/NETWORK accounts only — for a GROUP or admin-network account this (and get_dynamic_report_data) returns 403; use the classic get_campaign_breakdown_report for those instead. Returns the metamodel schema with all available dimensions, metrics, filters, and their allowed operators. You MUST call this tool BEFORE calling get_dynamic_report_data - do NOT guess dimension/metric/filter names or operators. The response tells you exactly which names are valid for columns and filters in the query. Per-conversion-rule metrics are shown compactly as a naming pattern + rule list; use name_filter to find a specific rule or narrow a large surface. WORKFLOW: 1) search_accounts to get account_id 2) Call THIS tool to discover available options 3) Use the returned names/operators to build a valid report_query 4) Call get_dynamic_report_data with that query.

get_item

ChatGPT
Get details for one item on a campaign. "Item" and "ad" refer to the same object.

list_campaigns

ChatGPT
List campaigns on a Realize account, one page per call. Page through with page / page_size (defaults 1/10, max 10). CONVERSION GOAL: a campaign's conversion_rules lists only rules set on the campaign itself; the goal is usually set on its campaign group. An empty conversion_rules with campaign_group_id set does NOT mean no goal — call get_campaign_groups with campaign_group_id set to that id (see get_campaign).

list_cta_types

ChatGPT
List allowed values for cta.cta_type on create_native_item / update_native_item. The set is curated by Realize and changes over time; prefer this tool over hard-coded enums.

list_items

ChatGPT
List all items on a campaign. "Item" and "ad" refer to the same object.

list_time_zones

ChatGPT
List valid IANA time zone names for activity_schedule.time_zone on create_campaign / update_campaign (CUSTOM mode). Output values are directly usable on the wire (e.g. "America/New_York", "US/Eastern", "Europe/London"). The literal "USER_TIMEZONE" ("User time" — schedule follows each viewer's own local time) is also valid but is NOT returned here.

search_accounts

ChatGPT
Search for accounts by numeric ID or text query. Use this first — every other tool's account_id parameter takes the value from the account_id field returned here. Each result includes the account's currency, country, and time_zone_name — use these to choose budget amounts in the right currency and to populate activity_schedule.time_zone. Response metadata includes Total (full match count across pages). Keep page_size constant across pages to avoid duplicate or missing results. Each result also carries a type: PARTNER (an advertiser account) or NETWORK (spans many advertiser accounts) both support reporting. Results may also include GROUP accounts and admin networks: the DYNAMIC report (get_dynamic_report_settings / get_dynamic_report_data) returns 403 on those, but the classic get_campaign_breakdown_report DOES serve them. So for a GROUP/admin-network account use get_campaign_breakdown_report; use the flexible dynamic report only for a PARTNER or NETWORK account. Do not retry or re-authenticate on that 403.

search_audiences

ChatGPT
Read the audiences on a Realize account (read-only). Omit audience_id to list every audience across all subtypes (first-party, custom, combined, lookalike, third-party). Each entry carries id, audience_name, data_type and is_archived; the IDs populate audiences_targeting.value[].value: [int] on create_campaign / update_campaign. Set audience_id (with audience_type, which routes the lookup) to fetch one audience in full. The listing does NOT include the writable detail — condition/effects/status for PIXEL, combined_audience_components for COMBINED — so read an audience this way before calling update_audience, then send only the fields you want to change. The by-id read covers every subtype update_audience can write. Both forms return the same {account_id, ..., values: [...]} shape; audience_id just yields a one-entry list.

search_contextual_segments

ChatGPT
Search contextual segments available for targeting on an account. Segment IDs returned here populate contextual_segments_targeting.value[].value: [int] on create_campaign / update_campaign.

search_conversion_rules

ChatGPT
DEPRECATED — renamed. Call get_conversion_rules instead; this name is removed after 2026-11-01. Identical behaviour: it also fetches one rule by id, and the listing returns one page per call (page / page_size, defaults 1/25).

search_geos

ChatGPT
Search valid country / region / dma / city / postal_code values for create_campaign and update_campaign geo targeting. dimension=countries (no country_code) returns countries. dimension=regions|dma|cities|postal_codes requires country_code (ISO-2, e.g. "US"). DMA is US-only. Response shape: {dimension, values: [{code, name}, ...]}. The code field is what country_targeting / region_country_targeting / dma_country_targeting / city_targeting / postal_code_targeting accept on create_campaign and update_campaign — name is the human-readable label only (e.g. for regions, code="CA", name="California"; pass "CA" to region_country_targeting, not "California"). Example — list US states: { "dimension": "regions", "country_code": "US" }

search_lookalike_audiences

ChatGPT
Search lookalike audiences (CRM / pixel / PBP) available for targeting on an account. rule_ids returned here populate lookalike_audience_targeting.value[].value[].rule_id on create_campaign / update_campaign. Optional country_code (ISO-2) narrows to audiences targeting one country.

search_publishers

ChatGPT
Search publishers an account is allowed to target. Each result's account_id (NOT the display name) populates publisher_targeting.value and publisher_bid_modifier.target on create_campaign / update_campaign. query is required: pass a name substring to filter, or '*' to list all. Use publisher_ids for direct lookup. Pagination via page / page_size (page_size capped at 50). Results are trimmed to {id, name, account_id, country, is_active}.

search_techno

ChatGPT
Search valid technology-targeting values for create_campaign and update_campaign os_targeting (sub_categories) and browser_targeting. Other techno enums (platforms, os_family, connection_type) are inlined in their schema descriptions as small fixed enums and don't require discovery. dimension=browsers takes no extra args. dimension=operating_system_versions requires os_family (e.g. "iOS", "Android") and returns values usable as sub_categories on os_targeting items. Example — list iOS versions: { "dimension": "operating_system_versions", "os_family": "iOS" }

update_audience

ChatGPT
Update an audience on a Realize account. Send ONLY the fields you want to change: the backend partial-merges, so every field you omit keeps its stored value. audience_type must match what the audience actually is — it selects the endpoint. RETIRING AN AUDIENCE — there is no delete tool because the Realize API has no audience delete. Archive instead, and the subtypes spell it differently: PIXEL -> status: "ARCHIVED" (or DISABLED to pause collection without archiving) every other -> is_archived: true Both reverse (status ACTIVE / is_archived false). Once a COMBINED audience is archived, archiving state is the only thing an update may change. An archived audience also stops being readable by id — confirm the archive from the plain search_audiences listing, which keeps showing it with is_archived true. CUSTOMER_LIST TAKES A DIFFERENT, MUCH SMALLER BODY ON UPDATE than it does on create: only audience_name, audience_description, exclude_from_campaigns, is_archived and country. Its crm_targeting_type and crm_file_type are fixed at creation. Nothing here can change who is in an audience — membership is never editable through this tool for any subtype. To change a customer list's members, create a new one from a new file; there is no way to add or remove rows on an existing one. Read the audience first with search_audiences(audience_id=..., audience_type=...) to see its current values, then send just your edits — the read payload is not a valid update body. Do not echo it back: its read-only fields (id, advertiser_id, audience_size, provider, data_type, size, last_modified_by, last_modified_at) have no property here and are rejected as unknown parameters, as are fields belonging to the other audience_type. That is deliberate: an update usually carries one or two fields, so a typo is the whole edit, and a silently dropped field returns success having changed nothing. PIXEL: type, category and event_name are immutable — resending one unchanged is fine, changing one returns a READONLY error. COMBINED: combined_audience_components is FULL-REPLACE — the array you send becomes the whole definition, so read it, modify locally, and send the merged result. Omit it to leave the definition untouched.

update_campaign

ChatGPT
Update an existing campaign: scalars and any targeting block in one call. Prerequisites: call list_campaigns or get_campaign to obtain campaign_id and the owning account_id (the campaign's advertiser_id — a NETWORK/parent account can list the campaign but cannot update it), or use the id returned by create_campaign. Numeric account IDs are rejected. Targeting blocks are FULL-REPLACE: sending audiences_targeting with one audience wipes the others; sending country_targeting replaces all country targeting. To edit a single entry, read with get_campaign, modify locally, send the merged result. Scalar fields (name, cpc, is_active, …) are partial-merge — omitted scalars keep their prior value. Geo: send any subset of country_targeting, region_country_targeting, dma_country_targeting, city_targeting, postal_code_targeting. Each replaces only its own dimension. Sub-dimension mutex: at most one of region/dma/city/postal_code per campaign; INCLUDE country must already be set before a sub-dimension is accepted. All amounts are numbers in the account's default currency. Conditional rules (apply only when the gating field is in this request): - spending_limit_model = MONTHLY|ENTIRE → also send spending_limit. NONE → also send daily_cap. - daily_ad_delivery_model = STRICT → also send daily_cap. BALANCED → omit daily_cap. - bid_strategy = MAX_CONVERSIONS → cpc_cap allowed. Other strategies → omit cpc_cap. - bid_strategy = TARGET_CPA → also send cpa_goal. - pricing_model = VCPM → bid_strategy MUST be FIXED; cpc carries the vCPM rate; omit cpc_cap and cpa_goal; traffic_allocation_mode is forced to OPTIMIZED server-side. - If both start_date and end_date sent: end_date >= start_date. - Solo updates of a partner field (e.g. only spending_limit, or only cpa_goal) are allowed — stored gating field is reused. At least one updatable field besides account_id and campaign_id must be sent. Server-side constraints (returns 4xx if violated): - Some marketing_objective transitions are rejected mid-flight. - MOBILE_APP_INSTALL switching requires app fields not exposed here. - TERMINATED campaigns cannot be reactivated. - Lookalike: account must have user-segments edit permission and campaign must allow retargeting. Discovery: same as create_campaign — search_geos, search_techno, search_audiences, search_lookalike_audiences, search_contextual_segments, search_publishers, get_conversion_rules, list_time_zones. Each schema property below names which tool populates it. TIMEZONE — user's OWN / "my" / "local" / "user" / "viewer" time: set activity_schedule.time_zone to the literal string "USER_TIMEZONE". Do NOT run a shell command, read the host clock, call list_time_zones, or resolve it to an IANA name. Only use an IANA name when the user names a specific place (e.g. "New York time"). CORRECT: "Saturday 9am-9pm in my local timezone" -> "time_zone": "USER_TIMEZONE" INCORRECT: (detect host TZ, then) -> "time_zone": "America/New_York" Field shapes are identical to create_campaign — see its comprehensive example for every targeting block. Examples below focus on update-only patterns (partial-merge, classic-geo, full-replace within a section). Read-only — NEVER send: id, advertiser_id, status, approval_state, spent, policy_review, target_cpa, target_cpa_learning_status. All updates (including audiences, lookalike, contextual segments) go in one atomic call. Example — activate paused campaign and add publisher bid modifier (partial-merge): { "account_id": "acme-inc", "campaign_id": "49184816", "is_active": true, "publisher_bid_modifier": {"values": [{"target": "pub_alpha", "cpc_modification": 1.25}]} } Example — clear audience targeting only (other targeting untouched, full-replace within section): { "account_id": "acme-inc", "campaign_id": "49184816", "audiences_targeting": {"state": "ALL", "value": []} } Example — edit one classic geo dimension: { "account_id": "acme-inc", "campaign_id": "49184816", "region_country_targeting": {"type": "INCLUDE", "value": ["CA", "NY"]} }

update_campaign_group

ChatGPT
Update a campaign group. Send only the fields to change; everything omitted keeps its current value (the tool re-sends the attribution windows and budget schedule that the API would otherwise clear). Read the group first with get_campaign_groups to see its values. The kpi must stay one the group's objective allows — changing the objective alone fails if the current kpi is not offered for the new one, so send both: BRAND_AWARENESS: CTR, CPC, CPM DRIVE_WEBSITE_TRAFFIC: CTR, CPC, CPQV LEADS_GENERATION: CTR, CPC, CPA, CPL, ROAS ONLINE_PURCHASES: ROAS, CPA, CTR, CPC MOBILE_APP_INSTALL: CPI, ROAS, CPA, CTR, CPC Pause or resume the whole group with is_active. Changing budget_distribution_method or the budget of a shared group updates every campaign in it. A REALIZE_PLUS group cannot change marketing_objective, kpi, kpi_data or budget_distribution_method. There is no delete here: deleting a group terminates all its campaigns — use is_active=false.

update_conversion_rule

ChatGPT
Update a conversion rule on a Realize account. Send ONLY the fields you want to change: the backend partial-merges, so every field you omit — including condition and effects — keeps its stored value. Do not echo back an object from get_conversion_rules: its explicit nulls (view_through_look_back_window, event_name, the condition members) fail input validation, and its fields with no property here (id, advertiser_id, pixel_id, exclude_from_campaigns, external_id, partner, tracked_elements) are rejected as unknown parameters. Unknown field names are rejected rather than ignored, so a misspelling fails loudly instead of dropping your edit and reporting success. Read a rule to see its current values, then send just your edits. type/category/event_name are immutable — resending one unchanged is accepted, changing one returns a READONLY error (which may name eventName even when type was the field you changed). Retire a rule by setting status to DISABLED/ARCHIVED. Pixel-based (DSP) rules can be renamed and retired here, but their pixel binding and other DSP-only fields are editable only in the Realize console.

update_display_item

ChatGPT
Update an existing display item: top-level scalars, the display_data block, and any nested block (verification_pixel, viewability_tag) in one call. "Item" and "ad" refer to the same object — this tool updates one. Prerequisites: call list_campaigns or get_campaign (campaign_id plus the owning account_id = the campaign's advertiser_id, not a NETWORK/parent account), then list_items or get_item (item_id). Numeric account IDs are rejected. Partial-merge for scalars: fields omitted from the request keep their prior value. The creative discriminator on update follows the same mutex as create — send at most one of ad_tag or asset_url: - ad_tag (3P) swaps the tag. dimensions is required when ad_tag is sent and replaces the prior dimension. - asset_url (1P) replaces the hosted asset; Realize re-ingests by URL file extension. Do NOT send dimensions with asset_url. Items created in one mode generally cannot be flipped to the other; the server rejects mode-switch attempts with a 4xx. Array semantics — FULL-REPLACE within section: sending verification_pixel or viewability_tag overwrites the entire prior list for that field. Send [] to clear. To edit one entry, read with get_item, modify locally, send the merged result. Editability: items in CRAWLING / PENDING_APPROVAL accept full edits. Items in RUNNING / PAUSED practically only accept is_active toggles and minor metadata; the API surfaces 4xx for non-allowed transitions. REJECTED items cannot be edited — recreate. Scope: display items only — for editing native image ads, carousels and performance videos use update_native_item. HOSTED_SOCIAL, RSS feed items, and the Creative Library are not supported. Not applicable to display creatives: branding_text and cta — the ad tag or hosted asset handles its own branding and click; sending these has no effect. Read-only — NEVER send: id, campaign_id (set via path), creative_type, display_data.display_ad_type, status, approval_state, learning_state, policy_review. At least one updatable field besides account_id, campaign_id, and item_id must be sent. All updates (including verification_pixel and viewability_tag) go in one atomic call. Comprehensive examples (only account_id, campaign_id, and item_id are mandatory, plus at least one updatable field). { "account_id": "acme-inc", "campaign_id": "49184816", "item_id": "987654321", "url": "https://example.com/landing-v2", "ad_tag": "<ins class=\"dcmads\" style=\"display:inline-block;width:300px;height:250px\" data-dcm-placement=\"N12345.1234567ACME/B12345678.123456789\" data-dcm-rendering-mode=\"script\" data-dcm-https-only data-dcm-resettable-device-id=\"\" data-dcm-app-id=\"\" data-dcm-click-tracker=\"${CLICK_URL}\"><script src=\"https://www.googletagservices.com/dcm/dcmads.js\"></script></ins>", "dimensions": [{"width": 300, "height": 250}], "creative_name": "Acme Spring Sale 300x250 — v2", "is_active": true, "verification_pixel": { "verification_pixel_items": [ {"url": "https://verifier.example.com/c?x=1", "verification_pixel_type": "CLICK"}, {"url": "https://verifier.example.com/v?x=1", "verification_pixel_type": "VIEWABLE_IMPRESSION"} ] }, "viewability_tag": { "values": [ {"tag": "<noscript class=...></noscript><script src=...></script>", "type": "IAS"} ] } } Example — pause display item (partial-merge, scalars only): { "account_id": "acme-inc", "campaign_id": "49184816", "item_id": "987654321", "is_active": false } Example — swap ad tag without touching dimensions (display_data partial-merge): { "account_id": "acme-inc", "campaign_id": "49184816", "item_id": "987654321", "ad_tag": "<script>...new tag...</script>" } Example — swap the hosted asset on a 1P item (re-ingest from new URL): { "account_id": "acme-inc", "campaign_id": "49184816", "item_id": "987654321", "asset_url": "https://cdn.example.com/creatives/acme-spring-300x250-v2.png" } Example — clear all verification pixels (full-replace within section): { "account_id": "acme-inc", "campaign_id": "49184816", "item_id": "987654321", "verification_pixel": {"verification_pixel_items": []} }

update_native_item

ChatGPT
Update an existing native item: scalars and any nested block (cta, verification_pixel, viewability_tag) in one call. "Item" and "ad" refer to the same object — this tool updates one. Prerequisites: call list_campaigns or get_campaign (campaign_id plus the owning account_id = the campaign's advertiser_id, not a NETWORK/parent account), then list_items or get_item (item_id). Numeric account IDs are rejected. Partial-merge for scalars: fields omitted from the request keep their prior value. Array semantics — FULL-REPLACE within section: sending verification_pixel or viewability_tag overwrites the entire prior list for that field. Send [] to clear. To edit one entry, read with get_item, modify locally, send the merged result. Editability: items in PENDING_APPROVAL accept full edits. Items in RUNNING / PAUSED practically only accept is_active toggles and minor metadata; the API surfaces 4xx for non-allowed transitions. REJECTED items cannot be edited — recreate. Scope: native image ads, carousels and performance videos — for editing third-party display ads or 1P hosted display assets use update_display_item. RSS feed items and the Creative Library are not supported. Performance video items: video_url replaces the video (the tool reads the item first and refuses anything that is not a video ad). On a video ad, thumbnail_url is the fallback image shown where the video cannot play; replacing the video does not regenerate it, so send a matching thumbnail_url with the new video_url. A link Backstage cannot use fails with a 400 ("URL is not accessible" / "Invalid video_url") — fix the link, don't retry. Carousel cards (cards, remove_cards, fallback_policy; carousel items only): Backstage treats the card list as FULL-REPLACE and deletes any current card left out of it. So call get_item first, then send cards with EVERY current card's id — {"id": "<id>"} alone keeps a card unchanged, add only the fields you change — plus any new cards without an id. To delete cards, name them in remove_cards (it can be sent alone). The tool reads the carousel before sending and refuses an update that would drop a card you did not name in remove_cards, or leave fewer than 3 active cards or more than 10 cards. Prefer pausing a card (is_active: false) to removing it: removal cannot be undone. Omit all three to edit only the carousel's own fields; its cards stay untouched. Discovery (call before constructing the payload to resolve IDs/names): - list_cta_types → cta.cta_type allowed values. Read-only — NEVER send: id, campaign_id (set via path), type, status, approval_state, learning_state, orientation, policy_review. At least one updatable field besides account_id, campaign_id, and item_id must be sent. All updates (including verification_pixel and viewability_tag) go in one atomic call. Comprehensive example (every available field set; trim what you don't need — only account_id, campaign_id, and item_id are mandatory, plus at least one updatable field). Plain English: refresh an existing ad — new headline / body / thumbnail / branding, Learn More CTA, CLICK + VIEWABLE_IMPRESSION verification pixels and a MOAT viewability tag, shipped active. { "account_id": "acme-inc", "campaign_id": "49184816", "item_id": "987654321", "url": "https://example.com/landing", "title": "Save 25% — Extended", "description": "Spring sale extended through end of June.", "thumbnail_url": "https://cdn.example.com/spring-v2.jpg", "branding_text": "Acme", "creative_name": "Acme Spring Sale Native — v2", "cta": {"cta_type": "LEARN_MORE"}, "is_active": true, "verification_pixel": { "verification_pixel_items": [ {"url": "https://verifier.example.com/c?x=1", "verification_pixel_type": "CLICK"}, {"url": "https://verifier.example.com/v?x=1", "verification_pixel_type": "VIEWABLE_IMPRESSION"} ] }, "viewability_tag": { "values": [ {"tag": "<noscript class=...></noscript><script src=...></script>", "type": "IAS"} ] } } Example — pause item (partial-merge, scalars only): { "account_id": "acme-inc", "campaign_id": "49184816", "item_id": "987654321", "…

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/realize-plugin"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/realize-plugin)