diff --git a/models/llm_negotiation_market/README.md b/models/llm_negotiation_market/README.md
new file mode 100644
index 0000000..d062088
--- /dev/null
+++ b/models/llm_negotiation_market/README.md
@@ -0,0 +1,38 @@
+## How to run
+# Option 1: Google Colab (recommended)
+Open model.ipynb in Google Colab
+https://colab.research.google.com/drive/1yLOkKcxm4rOBbcFjULn_gVbCnVYwaoBP?usp=sharing
+
+# Option 2: Local (requires Linux/WSL due to Windows DLL policy)
+pip install mesa mesa-llm groq litellm
+python model.py
+
+
+# LLM Negotiation Market Model
+
+A Mesa-LLM model where buyer and seller agents use Chain-of-Thought 
+reasoning to negotiate prices over multiple steps.
+
+## What I learned
+
+- `LLMAgent.step()` is empty by default — you must call 
+  `self.reasoning.plan()` manually to trigger LLM thinking
+- `CoTReasoning` makes two LLM calls: one for thinking, one for 
+  tool execution. When no tools are registered, the second call 
+  crashes with a Groq 400 error (`tool_choice="required"` with empty tools)
+- Fixed by subclassing `CoTReasoning` into `SimpleCoTReasoning` that 
+  skips the tool execution step — only uses the thinking/planning call
+- Agents have real memory across steps — in Step 2, agents referenced 
+  their Step 1 decisions without any extra code
+- Three reasoning engines available: `CoTReasoning`, `ReActReasoning`, 
+  `ReWOOReasoning`
+- `llm_model` needs `"groq/"` prefix for LiteLLM routing
+
+## How to run
+pip install mesa mesa-llm groq litellm
+export GROQ_API_KEY=your_key
+python model.py
+
+## Potential Issues Found in mesa-llm
+- `CoTReasoning` crashes when no tools registered (tool_choice="required" bug)
+  → This could be a good issue to open on mesa/mesa-llm!
\ No newline at end of file
diff --git a/models/llm_negotiation_market/model.py b/models/llm_negotiation_market/model.py
new file mode 100644
index 0000000..ab32167
--- /dev/null
+++ b/models/llm_negotiation_market/model.py
@@ -0,0 +1,99 @@
+import os
+import mesa
+from mesa_llm.llm_agent import LLMAgent
+from mesa_llm.reasoning.cot import CoTReasoning
+
+os.environ["GROQ_API_KEY"] = "your_groq_api_key_here"
+
+
+class SimpleCoTReasoning(CoTReasoning):
+    """
+    Subclass of CoTReasoning that skips tool execution.
+    Only uses the first LLM call (the thinking/planning step).
+    This avoids the tool_choice='required' error when no tools are registered.
+    """
+
+    def plan(self, prompt=None, obs=None, ttl=1, selected_tools=None):
+        if prompt is None:
+            if self.agent.step_prompt is not None:
+                prompt = self.agent.step_prompt
+            else:
+                raise ValueError("No prompt provided and agent.step_prompt is None.")
+
+        if obs is None:
+            obs = self.agent.generate_obs()
+
+        llm = self.agent.llm
+        obs_str = str(obs)
+
+        self.agent.memory.add_to_memory(
+            type="Observation", content={"content": obs_str}
+        )
+
+        system_prompt = self.get_cot_system_prompt(obs)
+        llm.system_prompt = system_prompt
+
+        rsp = llm.generate(
+            prompt=prompt,
+            tool_schema=None,
+            tool_choice="none",
+        )
+
+        chaining_message = rsp.choices[0].message.content
+        self.agent.memory.add_to_memory(
+            type="Plan", content={"content": chaining_message}
+        )
+        if hasattr(self.agent, "_step_display_data"):
+            self.agent._step_display_data["plan_content"] = chaining_message
+
+        return chaining_message
+
+
+class SellerAgent(LLMAgent):
+    def __init__(self, model, price):
+        super().__init__(
+            model=model,
+            reasoning=SimpleCoTReasoning,
+            llm_model="groq/llama-3.3-70b-versatile",
+            system_prompt="You are a seller in a market. Reply in one sentence.",
+            step_prompt=f"Your price is {price}. Should you lower it to attract buyers?",
+        )
+        self.price = price
+
+    def step(self):
+        response = self.reasoning.plan()
+        print(f"Seller {self.unique_id} (price={self.price}): {response}")
+
+
+class BuyerAgent(LLMAgent):
+    def __init__(self, model, budget):
+        super().__init__(
+            model=model,
+            reasoning=SimpleCoTReasoning,
+            llm_model="groq/llama-3.3-70b-versatile",
+            system_prompt="You are a buyer in a market. Reply in one sentence.",
+            step_prompt=f"Your budget is {budget}. What price will you offer?",
+        )
+        self.budget = budget
+
+    def step(self):
+        response = self.reasoning.plan()
+        print(f"Buyer {self.unique_id} (budget={self.budget}): {response}")
+
+
+class NegotiationMarket(mesa.Model):
+    def __init__(self, n_buyers=2, n_sellers=2):
+        super().__init__()
+        for _ in range(n_sellers):
+            SellerAgent(self, price=100)
+        for _ in range(n_buyers):
+            BuyerAgent(self, budget=150)
+
+    def step(self):
+        self.agents.shuffle_do("step")
+
+
+model = NegotiationMarket()
+for i in range(2):
+    print(f"\n--- Step {i + 1} ---")
+    model.step()