diff --git a/src/comind.py b/src/comind.py
new file mode 100644
index 0000000..5c321b4
--- /dev/null
+++ b/src/comind.py
@@ -0,0 +1,122 @@
+# Code to manage cominds and prompts
+
+import os
+import re
+import json
+import src.structured_gen as sg
+from pydantic import BaseModel
+from src.lexicon_utils import lexicon_of, multiple_of_schema
+
+PROMPT_DIR = "prompts/cominds"
+COMMON_PROMPT_DIR = "prompts/common"
+
+class Comind(BaseModel):
+    name: str
+    prompt_path: str
+    common_prompt_dir: str
+
+    def load_prompt(self):
+        with open(self.prompt_path, "r") as f:
+            return f.read()
+        
+    def load_common_prompts(self):
+        common_prompts = {}
+        for file in os.listdir(self.common_prompt_dir):
+            with open(os.path.join(self.common_prompt_dir, file), "r") as f:
+                bn = os.path.basename(file).replace(".co", "")
+                common_prompts[bn] = f.read()
+        return common_prompts
+    
+    def to_prompt(self, context_dict: dict):
+        prompt = self.load_prompt()
+        common_prompts = self.load_common_prompts()
+
+        # Merge common prompts into context_dict
+        for common_prompt in common_prompts:
+            if common_prompt in context_dict:
+                print(f"Warning: Common prompt {common_prompt} already in context_dict. Common prompt names are reserved and should not be overridden.")
+            context_dict[common_prompt] = common_prompts[common_prompt]
+
+        # Format prompt
+        prompt = prompt.format(**context_dict)
+        return prompt
+    
+    def split_prompts(self, context_dict: dict):
+        """
+        Splits a co file into system, schema, and user messages.
+
+        The co file is expected to have the following format:
+        <CO|SYSTEM> ... </CO|SYSTEM>
+        <CO|SCHEMA> ... </CO|SCHEMA>
+        <CO|USER> ... </CO|USER>
+        
+        Tags may be in any order. User prompts are required.
+        """
+        prompt = self.to_prompt(context_dict)
+
+        # Load a prompt from a file and strip out the tags
+        user_regex = r"<CO\|USER>(.*?)</CO\|USER>"
+        system_regex = r"<CO\|SYSTEM>(.*?)</CO\|SYSTEM>"
+        schema_regex = r"<CO\|SCHEMA>(.*?)</CO\|SCHEMA>"
+
+        user_match = re.search(user_regex, prompt, re.DOTALL)
+        system_match = re.search(system_regex, prompt, re.DOTALL)
+        schema_match = re.search(schema_regex, prompt, re.DOTALL)
+
+        user_prompt = user_match.group(1).strip() if user_match else ""
+        system_prompt = system_match.group(1).strip() if system_match else ""
+        schema_prompt = schema_match.group(1).strip() if schema_match else ""
+
+        if not user_match:
+            raise ValueError("User prompt is required.")
+        
+        system_prompt = system_match.group(1) if system_match else None
+        schema_prompt = schema_match.group(1) if schema_match else None
+        user_prompt = user_match.group(1)
+
+        return {
+            "system": system_prompt.strip() if system_prompt else None,
+            "schema": schema_prompt.strip() if schema_prompt else None,
+            "user": user_prompt.strip()
+        }
+    
+    def messages(self, values: dict):
+        messages = []
+        if values["system"]:
+            messages.append({"role": "system", "content": values["system"]})
+        messages.append({"role": "user", "content": values["user"]})
+        return messages
+    
+    def run(self, context_dict: dict, schema: str = None):
+        prompts = self.split_prompts(context_dict)
+        messages = self.messages(prompts)
+
+        if not schema:
+            # Check if we have one in the prompts dict. Must be 
+            # nonzero length and valid JSON.
+            if "schema" in prompts:
+                try:
+                    json.loads(prompts["schema"])
+                    schema = prompts["schema"]
+                except json.JSONDecodeError:
+                    raise ValueError("Schema is not valid JSON.")
+            else:
+                raise ValueError("Schema is required.")
+
+        return sg.generate_by_schema(messages, schema)
+
+if __name__ == "__main__":
+    # Test the Comind class
+    comind = Comind(
+        common_prompt_dir="prompts/common/",
+        name="conceptualizer",
+        prompt_path="prompts/cominds/conceptualizer.co",
+    )
+    print(comind.load_prompt())
+    print(comind.load_common_prompts())
+
+    context_dict = {
+        "content": "Hello, world!"
+    }
+    print(comind.to_prompt(context_dict))
+    print(comind.run(context_dict, multiple_of_schema("concepts", lexicon_of("me.comind.blip.concept"))))
\ No newline at end of file