Menu de la documentation
Bien démarrer
Qu'est-ce qu'Opus GrowthDémarrage rapideAjoutez-le à votre IAFonctionnementSécurité et validationCréditsOffresFAQApplications IA
Toutes les applicationsChatGPTClaudeGrokPerplexityClaude CodeCodexGemini CLIOpenClawCursorVS CodeWindsurfAutres MCPPlateformes publicitaires
Google AdsMeta AdsLinkedIn AdsMicrosoft AdvertisingTikTok AdsChatGPT AdsAgences et MCCSources de données
Search ConsoleGoogle AnalyticsYouTubeProfil d'établissementTag ManagerSites et SEO
AI ServerWordPressSEO IntelligenceFraude au clicRéférence des outils
Tous les outilsGoogle AdsMeta AdsTikTok AdsLinkedIn AdsMicrosoft AdvertisingChatGPT AdsSearch ConsoleGoogle AnalyticsYouTubeProfil d'établissementTag ManagerWordPressAI ServerSEO IntelligenceFraude au clicMulticanalCompte et facturationLangue
Experiments (list / end)
Lists AND ends campaign experiments (A/B tests).
- Type
- Effectue des modifications
- Crédits
- 1
- Plateforme
- Google Ads
- Validation
- Aperçu d'abord, appliqué après validation
Ce qu'il fait
Lists AND ends campaign experiments (A/B tests). An experiment LOCKS the campaigns in it: while it runs, bidding strategy / budget / targeting writes on those campaigns are rejected by Google — so this tool is the way out of that dead-end.
action: • list — every experiment with id, name, status, type, dates AND the campaigns each one locks (read-only, no confirm). • end — stops a RUNNING experiment now (experiment_id required). Traffic goes back to the original campaign, the results collected so far are kept, and the campaigns are unlocked so you can change bids/budget/targeting again. This is what you want in almost every case. • promote — permanently APPLIES the experiment's changes to the original campaign, then ends it (experiment_id required). promote applies a winning test. • remove: deletes the experiment record entirely (experiment_id required). remove is for experiments that never ran or are no longer wanted; results are lost.
Statuses that lock a campaign: ENABLED, SETUP, INITIATED. HALTED / PROMOTED / GRADUATED / REMOVED do not lock. confirm=false → dry-run preview. To CREATE an experiment (traffic split, different landing page or ad text) use manage_experiment.
La description est en anglais car c'est le texte officiel que voit votre IA.
MCP: list_experiments
Comment l'utiliser
Vous n'appelez pas les outils par leur nom. Écrivez ce que vous voulez dans ChatGPT, Claude ou un autre assistant connecté à Opus Growth ; l'IA choisit cet outil quand il convient.
Paramètres
| Nom | Type | Obligatoire | Par défaut | |
|---|---|---|---|---|
| customer_id | string | Obligatoire | · | Google Ads customer account ID to query or change. Use the numeric ID returned by list_accounts; omit separators unless this tool explicitly says otherwise. |
| action | string | Facultatif | list | list · end · promote · remove · Requested action for the selected resource; choose one of the actions allowed by this schema. |
| experiment_id | string | Facultatif | · | Exact experiment identifier to target. Use the ID returned by an Opus Growth list or create operation, not a display name. |
| confirm | boolean | Facultatif | false | Safety confirmation for a state-changing operation. Leave false to preview or validate; set true only after the user explicitly approves the shown change. |
Outils associés
account_auditAccount Snapshotaccount_snapshotAdd Keywordsadd_keywordsAdd Negative Keywordadd_negative_keywordAdd PMax Imagesadd_pmax_imagesApply Recommendationapply_recommendationBudget Pacing Alertsbudget_pacing_alertsCompetitor Adscompetitor_adsConnect Business Profileconnect_business_profile