diff --git a/enterprise/lib/enterprise/captain/base_task_service.rb b/enterprise/lib/enterprise/captain/base_task_service.rb
index b080ff807..bda4708af 100644
--- a/enterprise/lib/enterprise/captain/base_task_service.rb
+++ b/enterprise/lib/enterprise/captain/base_task_service.rb
@@ -28,7 +28,7 @@ module Enterprise::Captain::BaseTaskService
   end
 
   def increment_usage
-    credits = Llm::Models.credit_multiplier_for(configured_model)
+    credits = using_hook_key? ? 1 : Llm::Models.credit_multiplier_for(configured_model)
     Rails.logger.info("[CAPTAIN][#{self.class.name}] Incrementing response usage for account #{account.id} by #{credits} credits")
     account.increment_response_usage(credits: credits)
   end
diff --git a/lib/captain/base_task_service.rb b/lib/captain/base_task_service.rb
index 739239cd2..e14048451 100644
--- a/lib/captain/base_task_service.rb
+++ b/lib/captain/base_task_service.rb
@@ -72,6 +72,11 @@ class Captain::BaseTaskService
       build_ruby_llm_response(response, messages)
     end
   rescue StandardError => e
+    # TODO: RubyLLM throws the RubyLLM::Error when any API based error occurs
+    # Just like inboxes have a reauthroizable error handling mechanism,
+    # we should handle the errors if the account is using their own key.
+    # If more than 5 errors occur, we disable the hook and notify the admins
+    # see: https://rubyllm.com/error-handling/#rubyllm-error-hierarchy
     ChatwootExceptionTracker.new(e, account: account).capture_exception
     { error: e.message, request_messages: messages }
   end
@@ -128,6 +133,8 @@ class Captain::BaseTaskService
   end
 
   def api_key
+    # TODO: Once we support multiple LLM providers, ensure provider key is correctly looked up
+    # Currently works because the only model/provider is OpenAI
     @api_key ||= openai_hook&.settings&.dig('api_key') || system_api_key
   end
 
@@ -135,6 +142,12 @@ class Captain::BaseTaskService
     @openai_hook ||= account.hooks.find_by(app_id: 'openai', status: 'enabled')
   end
 
+  def using_hook_key?
+    # TODO: Once we support multiple LLM providers, ensure provider key is correctly looked up
+    # Currently works because the only model/provider is OpenAI
+    openai_hook&.settings&.dig('api_key').present?
+  end
+
   def system_api_key
     @system_api_key ||= InstallationConfig.find_by(name: 'CAPTAIN_OPEN_AI_API_KEY')&.value
   end
diff --git a/spec/enterprise/lib/captain/base_task_service_spec.rb b/spec/enterprise/lib/captain/base_task_service_spec.rb
index e0dd5fbab..cb9679bb5 100644
--- a/spec/enterprise/lib/captain/base_task_service_spec.rb
+++ b/spec/enterprise/lib/captain/base_task_service_spec.rb
@@ -69,6 +69,22 @@ RSpec.describe Captain::BaseTaskService, type: :model do
       service.perform
     end
 
+    context 'when account uses their own OpenAI key via hook' do
+      let!(:openai_hook) { create(:integrations_hook, :openai, account: account, settings: { 'api_key' => 'user-own-key' }) }
+
+      it 'uses 1 credit regardless of model cost' do
+        allow(account).to receive(:captain_editor_model).and_return('gpt-5.1')
+        expect(account).to receive(:increment_response_usage).with(credits: 1)
+        service.perform
+      end
+
+      it 'uses 1 credit even for expensive models' do
+        allow(account).to receive(:captain_editor_model).and_return('gpt-4.1')
+        expect(account).to receive(:increment_response_usage).with(credits: 1)
+        service.perform
+      end
+    end
+
     context 'when result has an error' do
       let(:perform_result) { { error: 'API Error' } }