diff --git a/sentry-ruby/Gemfile b/sentry-ruby/Gemfile
index 68dc0d2b4..6d68e7706 100644
--- a/sentry-ruby/Gemfile
+++ b/sentry-ruby/Gemfile
@@ -35,6 +35,7 @@ gem "webrick"
 gem "faraday"
 gem "excon"
 gem "webmock"
+gem "ruby_llm"
 
 group :sequel do
   gem "sequel"
diff --git a/sentry-ruby/lib/sentry/ruby_llm.rb b/sentry-ruby/lib/sentry/ruby_llm.rb
new file mode 100644
index 000000000..3b30a71ad
--- /dev/null
+++ b/sentry-ruby/lib/sentry/ruby_llm.rb
@@ -0,0 +1,97 @@
+# frozen_string_literal: true
+
+module Sentry
+  module RubyLLM
+    OP_NAME = "gen_ai.chat"
+    EXECUTE_TOOL_OP_NAME = "gen_ai.execute_tool"
+    SPAN_ORIGIN = "auto.gen_ai.ruby_llm"
+    LOGGER_NAME = :ruby_llm_logger
+
+    module Patch
+      def ask(message = nil, with: nil, &block)
+        return super unless Sentry.initialized?
+
+        Sentry.with_child_span(op: OP_NAME, start_timestamp: Sentry.utc_now.to_f, origin: SPAN_ORIGIN) do |span|
+          result = super
+          model_id = @model&.id
+
+          if span
+            span.set_description("chat #{model_id}")
+            span.set_data("gen_ai.operation.name", "chat")
+            span.set_data("gen_ai.request.model", model_id)
+            span.set_data("gen_ai.system", @model&.provider)
+
+            if (response = @messages&.last)
+              span.set_data("gen_ai.response.model", response.model_id) if response.respond_to?(:model_id)
+
+              if response.respond_to?(:input_tokens) && response.input_tokens
+                span.set_data("gen_ai.usage.input_tokens", response.input_tokens)
+              end
+
+              if response.respond_to?(:output_tokens) && response.output_tokens
+                span.set_data("gen_ai.usage.output_tokens", response.output_tokens)
+              end
+            end
+
+            if instance_variable_defined?(:@temperature) && @temperature
+              span.set_data("gen_ai.request.temperature", @temperature)
+            end
+          end
+
+          record_breadcrumb("chat", model_id, @model&.provider)
+
+          result
+        end
+      end
+
+      def execute_tool(tool_call)
+        return super unless Sentry.initialized?
+
+        Sentry.with_child_span(op: EXECUTE_TOOL_OP_NAME, start_timestamp: Sentry.utc_now.to_f, origin: SPAN_ORIGIN) do |span|
+          result = super
+
+          if span
+            span.set_description("execute_tool #{tool_call.name}")
+            span.set_data("gen_ai.operation.name", "execute_tool")
+            span.set_data("gen_ai.tool.name", tool_call.name)
+            span.set_data("gen_ai.tool.call.id", tool_call.id)
+            span.set_data("gen_ai.tool.type", "function")
+
+            if Sentry.configuration.send_default_pii
+              span.set_data("gen_ai.tool.call.arguments", tool_call.arguments.to_json) if tool_call.arguments
+              span.set_data("gen_ai.tool.call.result", result.to_s[0..499]) if result
+            end
+          end
+
+          result
+        end
+      end
+
+      private
+
+      def record_breadcrumb(operation, name, provider = nil)
+        return unless Sentry.initialized?
+        return unless Sentry.configuration.breadcrumbs_logger.include?(LOGGER_NAME)
+
+        Sentry.add_breadcrumb(
+          Sentry::Breadcrumb.new(
+            level: :info,
+            category: OP_NAME,
+            type: :info,
+            data: {
+              operation: operation,
+              name: name,
+              provider: provider
+            }.compact
+          )
+        )
+      end
+    end
+  end
+end
+
+Sentry.register_patch(:ruby_llm) do
+  if defined?(::RubyLLM::Chat)
+    ::RubyLLM::Chat.prepend(Sentry::RubyLLM::Patch)
+  end
+end
diff --git a/sentry-ruby/spec/sentry/ruby_llm_spec.rb b/sentry-ruby/spec/sentry/ruby_llm_spec.rb
new file mode 100644
index 000000000..c8a55c2c6
--- /dev/null
+++ b/sentry-ruby/spec/sentry/ruby_llm_spec.rb
@@ -0,0 +1,226 @@
+# frozen_string_literal: true
+
+require "spec_helper"
+
+# Stub RubyLLM classes before loading the patch
+module RubyLLM
+  class Model
+    attr_accessor :id, :provider
+
+    def initialize(id:, provider:)
+      @id = id
+      @provider = provider
+    end
+  end
+
+  class Message
+    attr_accessor :model_id, :input_tokens, :output_tokens, :role, :content
+
+    def initialize(role:, content:, model_id: nil, input_tokens: nil, output_tokens: nil)
+      @role = role
+      @content = content
+      @model_id = model_id
+      @input_tokens = input_tokens
+      @output_tokens = output_tokens
+    end
+  end
+
+  class ToolCall
+    attr_accessor :name, :id, :arguments
+
+    def initialize(name:, id:, arguments: nil)
+      @name = name
+      @id = id
+      @arguments = arguments
+    end
+  end
+
+  class Chat
+    attr_reader :model, :messages
+
+    def initialize(model:)
+      @model = model
+      @messages = []
+    end
+
+    def ask(message = nil, with: nil, &block)
+      response = Message.new(
+        role: :assistant,
+        content: "Hello!",
+        model_id: @model.id,
+        input_tokens: 10,
+        output_tokens: 20
+      )
+      @messages << response
+      response
+    end
+
+    def execute_tool(tool_call)
+      "tool_result"
+    end
+  end
+end
+
+# Load the patch after stubs are defined
+require "sentry/ruby_llm"
+
+RSpec.describe Sentry::RubyLLM do
+  let(:model) { RubyLLM::Model.new(id: "gpt-4", provider: "openai") }
+  let(:chat) { RubyLLM::Chat.new(model: model) }
+
+  context "with tracing enabled" do
+    before do
+      perform_basic_setup do |config|
+        config.traces_sample_rate = 1.0
+        config.enabled_patches << :ruby_llm
+      end
+    end
+
+    it "records a span for ask" do
+      transaction = Sentry.start_transaction
+      Sentry.get_current_scope.set_span(transaction)
+
+      chat.ask("Hello")
+
+      spans = transaction.span_recorder.spans
+      ai_span = spans.find { |span| span.op == "gen_ai.chat" }
+
+      expect(ai_span).not_to be_nil
+      expect(ai_span.description).to eq("chat gpt-4")
+      expect(ai_span.origin).to eq("auto.gen_ai.ruby_llm")
+      expect(ai_span.data["gen_ai.operation.name"]).to eq("chat")
+      expect(ai_span.data["gen_ai.request.model"]).to eq("gpt-4")
+      expect(ai_span.data["gen_ai.system"]).to eq("openai")
+    end
+
+    it "records response data from the last message" do
+      transaction = Sentry.start_transaction
+      Sentry.get_current_scope.set_span(transaction)
+
+      chat.ask("Hello")
+
+      spans = transaction.span_recorder.spans
+      ai_span = spans.find { |span| span.op == "gen_ai.chat" }
+
+      expect(ai_span.data["gen_ai.response.model"]).to eq("gpt-4")
+      expect(ai_span.data["gen_ai.usage.input_tokens"]).to eq(10)
+      expect(ai_span.data["gen_ai.usage.output_tokens"]).to eq(20)
+    end
+
+    it "records a span for execute_tool" do
+      transaction = Sentry.start_transaction
+      Sentry.get_current_scope.set_span(transaction)
+
+      tool_call = RubyLLM::ToolCall.new(name: "get_weather", id: "call_123", arguments: { location: "Tokyo" })
+      chat.execute_tool(tool_call)
+
+      spans = transaction.span_recorder.spans
+      tool_span = spans.find { |span| span.op == "gen_ai.execute_tool" }
+
+      expect(tool_span).not_to be_nil
+      expect(tool_span.description).to eq("execute_tool get_weather")
+      expect(tool_span.origin).to eq("auto.gen_ai.ruby_llm")
+      expect(tool_span.data["gen_ai.operation.name"]).to eq("execute_tool")
+      expect(tool_span.data["gen_ai.tool.name"]).to eq("get_weather")
+      expect(tool_span.data["gen_ai.tool.call.id"]).to eq("call_123")
+      expect(tool_span.data["gen_ai.tool.type"]).to eq("function")
+    end
+
+    context "when send_default_pii is true" do
+      before { Sentry.configuration.send_default_pii = true }
+
+      it "records tool arguments and result" do
+        transaction = Sentry.start_transaction
+        Sentry.get_current_scope.set_span(transaction)
+
+        tool_call = RubyLLM::ToolCall.new(name: "get_weather", id: "call_123", arguments: { location: "Tokyo" })
+        chat.execute_tool(tool_call)
+
+        spans = transaction.span_recorder.spans
+        tool_span = spans.find { |span| span.op == "gen_ai.execute_tool" }
+
+        expect(tool_span.data["gen_ai.tool.call.arguments"]).to eq({ location: "Tokyo" }.to_json)
+        expect(tool_span.data["gen_ai.tool.call.result"]).to eq("tool_result")
+      end
+    end
+
+    context "when send_default_pii is false" do
+      before { Sentry.configuration.send_default_pii = false }
+
+      it "does not record tool arguments or result" do
+        transaction = Sentry.start_transaction
+        Sentry.get_current_scope.set_span(transaction)
+
+        tool_call = RubyLLM::ToolCall.new(name: "get_weather", id: "call_123", arguments: { location: "Tokyo" })
+        chat.execute_tool(tool_call)
+
+        spans = transaction.span_recorder.spans
+        tool_span = spans.find { |span| span.op == "gen_ai.execute_tool" }
+
+        expect(tool_span.data).not_to have_key("gen_ai.tool.call.arguments")
+        expect(tool_span.data).not_to have_key("gen_ai.tool.call.result")
+      end
+    end
+
+    it "sets correct timestamps on span" do
+      transaction = Sentry.start_transaction
+      Sentry.get_current_scope.set_span(transaction)
+
+      chat.ask("Hello")
+
+      spans = transaction.span_recorder.spans
+      ai_span = spans.find { |span| span.op == "gen_ai.chat" }
+
+      expect(ai_span.start_timestamp).not_to be_nil
+      expect(ai_span.timestamp).not_to be_nil
+      expect(ai_span.start_timestamp).to be < ai_span.timestamp
+    end
+  end
+
+  context "with breadcrumb logger enabled" do
+    before do
+      perform_basic_setup do |config|
+        config.traces_sample_rate = 1.0
+        config.enabled_patches << :ruby_llm
+        config.breadcrumbs_logger << :ruby_llm_logger
+      end
+    end
+
+    it "records a breadcrumb for ask" do
+      transaction = Sentry.start_transaction
+      Sentry.get_current_scope.set_span(transaction)
+
+      chat.ask("Hello")
+
+      llm_breadcrumb = Sentry.get_current_scope.breadcrumbs.peek
+
+      expect(llm_breadcrumb).not_to be_nil
+      expect(llm_breadcrumb.data[:operation]).to eq("chat")
+      expect(llm_breadcrumb.data[:name]).to eq("gpt-4")
+      expect(llm_breadcrumb.data[:provider]).to eq("openai")
+    end
+  end
+
+  context "without active transaction" do
+    before do
+      perform_basic_setup do |config|
+        config.traces_sample_rate = 1.0
+        config.enabled_patches << :ruby_llm
+      end
+    end
+
+    it "does not create spans when no transaction is active" do
+      result = chat.ask("Hello")
+      expect(result).to be_a(RubyLLM::Message)
+    end
+  end
+
+  context "when Sentry is not initialized" do
+    it "does not interfere with normal operations" do
+      # Create a fresh chat without Sentry initialized
+      fresh_chat = RubyLLM::Chat.new(model: model)
+      result = fresh_chat.ask("Hello")
+      expect(result).to be_a(RubyLLM::Message)
+    end
+  end
+end