diff --git a/aicodebot/cli.py b/aicodebot/cli.py
index 56758b9..0a2d517 100644
--- a/aicodebot/cli.py
+++ b/aicodebot/cli.py
@@ -1,7 +1,8 @@
 from aicodebot import version as aicodebot_version
 from aicodebot.coder import CREATIVE_TEMPERATURE, DEFAULT_MAX_TOKENS, Coder
-from aicodebot.config import get_config_file, read_config
+from aicodebot.config import get_config_file, get_local_data_dir, read_config
 from aicodebot.helpers import RichLiveCallbackHandler, create_and_write_file, exec_and_get_output, logger
+from aicodebot.learn import load_documents_from_repo, store_documents
 from aicodebot.prompts import DEFAULT_PERSONALITY, PERSONALITIES, generate_files_context, get_prompt
 from langchain.chains import LLMChain
 from langchain.memory import ConversationTokenBufferMemory
@@ -352,6 +353,36 @@ def fun_fact(verbose, response_token_size):
         chain.run(f"programming and artificial intelligence in the year {year}")
 
 
+@cli.command
+@click.option("-v", "--verbose", count=True)
+@click.option("-r", "--repo-url", help="The URL of the repository to learn from")
+def learn(repo_url, verbose):
+    """Learn new skills and gain additional knowledge from a repository"""
+    # Clone the supplied repo locally and walk through it, load it into a
+    # local vector store, and pre-query this vector store for the LLM to use a
+    # context for the prompt
+
+    setup_config()
+
+    owner, repo_name = Coder.parse_github_url(repo_url)
+
+    start_time = datetime.datetime.utcnow()
+
+    local_data_dir = get_local_data_dir()
+
+    Coder.clone_repo(repo_url, local_data_dir / "repos" / repo_name)
+    console.print("✅ Repo cloned.")
+
+    console.log("Loading documents")
+    vector_store_dir = local_data_dir / "vector_stores" / repo_name
+    documents = load_documents_from_repo(local_data_dir / "repos" / repo_name)
+    console.print("✅ Repo loaded and indexed.")
+
+    with console.status("Storing the repo in the vector store", spinner=DEFAULT_SPINNER):
+        store_documents(documents, vector_store_dir)
+    console.print(f"✅ Repo loaded and indexed in {datetime.datetime.utcnow() - start_time} seconds.")
+
+
 @cli.command
 @click.option("-c", "--commit", help="The commit hash to review (otherwise look at [un]staged changes).")
 @click.option("-v", "--verbose", count=True)
@@ -478,6 +509,7 @@ def sidekick(request, verbose, response_token_size, files):
         with Live(Markdown(""), auto_refresh=True) as live:
             callback = RichLiveCallbackHandler(live, bot_style)
             llm.callbacks = [callback]  # a fresh callback handler for each question
+
             chain.run({"task": human_input, "context": context})
 
         if request:
@@ -497,6 +529,7 @@ def setup_config():
         configure.callback(openai_api_key=os.getenv("OPENAI_API_KEY"), verbose=0)
         sys.exit(0)
     else:
+        os.environ["OPENAI_API_KEY"] = existing_config["openai_api_key"]
         return existing_config
 
 
diff --git a/aicodebot/coder.py b/aicodebot/coder.py
index c05ab69..546cc72 100644
--- a/aicodebot/coder.py
+++ b/aicodebot/coder.py
@@ -3,7 +3,7 @@
 from langchain.chat_models import ChatOpenAI
 from openai.api_resources import engine
 from pathlib import Path
-import fnmatch, functools, openai, tiktoken
+import fnmatch, functools, openai, re, subprocess, tiktoken
 
 DEFAULT_MAX_TOKENS = 512
 PRECISE_TEMPERATURE = 0.05
@@ -16,6 +16,18 @@ class Coder:
     git, and the local file system.
     """
 
+    @staticmethod
+    def clone_repo(repo_url, repo_dir):
+        """Clone a git repository to a directory."""
+        if Path(repo_dir).exists():
+            logger.info(f"Repo {repo_dir} already exists, updating it instead")
+            # Reset it first to make sure we don't have any local changes
+            subprocess.run(["git", "reset", "--hard"], cwd=repo_dir, check=True, stdout=subprocess.DEVNULL)
+            subprocess.run(["git", "pull"], cwd=repo_dir, check=True)
+        else:
+            logger.info(f"Cloning {repo_url} to {repo_dir}")
+            subprocess.run(["git", "clone", repo_url, repo_dir], check=True)
+
     @classmethod
     def generate_directory_structure(cls, path, ignore_patterns=None, use_gitignore=True, indent=0):
         """Generate a text representation of the directory structure of a path."""
@@ -198,3 +210,18 @@ def git_staged_files():
     @staticmethod
     def git_unstaged_files():
         return exec_and_get_output(["git", "diff", "HEAD", "--name-only"]).splitlines()
+
+    @staticmethod
+    def parse_github_url(repo_url):
+        """
+        Parse a GitHub URL and return the owner and repo name.
+        Returns: A tuple containing the owner and repo name.
+        """
+        pattern = r"(?:https:\/\/github\.com\/|git@github\.com:)([^\/]+)\/([^\/]+?)(?:\.git)?$"
+        match = re.match(pattern, repo_url)
+
+        if not match:
+            raise ValueError("URL is not a valid GitHub URL")
+
+        owner, repo = match.groups()
+        return owner, repo
diff --git a/aicodebot/config.py b/aicodebot/config.py
index 940fb0a..298dff2 100644
--- a/aicodebot/config.py
+++ b/aicodebot/config.py
@@ -3,6 +3,19 @@
 import os, yaml
 
 
+def get_local_data_dir():
+    data_dir = Path(os.getenv("AICODEBOT_LOCAL_DATA_DIR", str(Path.home() / ".aicodebot_data")))
+    # Make the directory if it doesn't exist
+    if not data_dir.exists():
+        logger.debug(f"Creating local data directory {data_dir}")
+        data_dir.mkdir()
+        # Create the subdirectories
+        (data_dir / "repos").mkdir()
+        (data_dir / "vector_stores").mkdir()
+
+    return data_dir
+
+
 def get_config_file():
     return Path(os.getenv("AICODEBOT_CONFIG_FILE", str(Path.home() / ".aicodebot.yaml")))
 
diff --git a/aicodebot/learn.py b/aicodebot/learn.py
new file mode 100644
index 0000000..b6d0adf
--- /dev/null
+++ b/aicodebot/learn.py
@@ -0,0 +1,125 @@
+from aicodebot.config import get_local_data_dir
+from aicodebot.helpers import logger
+from git import Repo
+from langchain.document_loaders import GitLoader, NotebookLoader
+from langchain.embeddings.openai import OpenAIEmbeddings
+from langchain.text_splitter import CharacterTextSplitter, Language, RecursiveCharacterTextSplitter
+from langchain.vectorstores import FAISS
+from pathlib import Path
+
+DEFAULT_EXCLUDE = [".csv", ".enex", ".json", ".jsonl"]
+
+
+def load_documents_from_repo(repo_dir, exclude=DEFAULT_EXCLUDE):
+    """Load a repo into the vector store."""
+
+    repo = Repo(repo_dir)
+    assert not repo.bare, f"Repo {repo_dir} does not appear to be a valid git repository"
+
+    # Check main first, then master, then give up
+    for branch in ["main", "master"]:
+        if branch in repo.heads:
+            default_branch = branch
+            break
+    else:
+        raise ValueError(f"Repo {repo_dir} does not have a main or master branch")
+
+    loader = GitLoader(repo_path=repo_dir, branch=default_branch)
+
+    documents = loader.load()
+    logger.info(f"Loaded {len(documents)} documents from {repo_dir}")
+
+    # Clean up
+    cleaned = []
+    logger.info("Cleaning up documents")
+    for document in documents:
+        content = document.page_content
+        if not content:
+            logger.debug(f"Skipping empty file {document.metadata['file_path']}")
+            continue
+
+        file_type = document.metadata["file_type"].lower()
+        if file_type in exclude:
+            logger.debug(f"Skipping excluded file {document.metadata['file_path']}")
+            continue
+
+        # Reload notebooks
+        if file_type == ".ipynb":
+            logger.debug(f"Reloading notebook {document.metadata['file_path']}")
+            new_document = NotebookLoader(repo_dir / document.metadata["file_path"]).load()[0]
+            # Use the original metadata, because it contains file_type
+            new_document.metadata = document.metadata
+            cleaned.append(new_document)
+        else:
+            cleaned.append(document)
+
+    return cleaned
+
+
+def store_documents(documents, vector_store_dir):
+    """Store documents in the vector store."""
+    vector_store_file = Path(vector_store_dir / "faiss_index")
+    embeddings = OpenAIEmbeddings()
+    if Path(vector_store_file).exists():
+        logger.info(f"Loading existing vector store {vector_store_file}")
+        return FAISS.load_local(vector_store_file, embeddings)
+
+    logger.info(f"Creating new vector store {vector_store_file}")
+
+    language_extension_map = {
+        ".py": Language.PYTHON,
+        ".ipynb": Language.PYTHON,
+        ".js": Language.JS,
+        ".ts": Language.JS,
+        ".html": Language.HTML,
+        ".md": Language.MARKDOWN,
+        ".mdx": Language.MARKDOWN,
+        ".go": Language.GO,
+        ".java": Language.JAVA,
+        ".c": Language.CPP,
+        ".cpp": Language.CPP,
+        ".php": Language.PHP,
+        ".rb": Language.RUBY,
+        ".xml": Language.HTML,
+    }
+
+    files = 0
+    chunks = []
+    for document in documents:
+        file_type = document.metadata["file_type"].lower()
+        files += 1
+
+        # Clean up
+        # Remove magic text that breaks processing
+        content = document.page_content.replace("<|end" + "of" + "text|>", "")  # noqa: ISC003
+
+        if file_type in language_extension_map:
+            # Use a recursive splitter for code files
+            logger.debug(
+                f"Processing {document.metadata['file_path']} as {language_extension_map[file_type].value} code"
+            )
+            splitter = RecursiveCharacterTextSplitter.from_language(
+                language=language_extension_map[document.metadata["file_type"].lower()], chunk_size=50, chunk_overlap=0
+            )
+        else:
+            # TODO: Check if it's a text file
+            if file_type not in [".txt", ".md", ".yml", ".yaml"]:
+                logger.info(f"Processing {document.metadata['file_path']} as a text file")
+            splitter = CharacterTextSplitter(separator="\n", chunk_size=1_000, chunk_overlap=200)
+
+        chunks += splitter.create_documents([content])
+
+    logger.info(f"Storing {len(chunks)} chunks from {files} files in {vector_store_dir}")
+    vector_store = FAISS.from_documents(chunks, embeddings)
+    vector_store.save_local(vector_store_file)
+    return vector_store
+
+
+def load_learned_repo(repo_name):
+    """Load a vector store from a learned repo."""
+    vector_store_file = Path(get_local_data_dir() / "vector_stores" / repo_name / "faiss_index")
+    if not vector_store_file.exists():
+        raise ValueError(f"Vector store for {repo_name} does not exist. Please run `aicodebot learn $githuburl` first.")
+
+    embeddings = OpenAIEmbeddings()
+    return FAISS.load_local(vector_store_file, embeddings)
diff --git a/requirements/requirements.in b/requirements/requirements.in
index 6bb3e16..7888e20 100644
--- a/requirements/requirements.in
+++ b/requirements/requirements.in
@@ -7,6 +7,7 @@
 
 beautifulsoup4 # needed by langchain
 click # command line interface helpers
+faiss-cpu
 GitPython
 langchain
 loguru
diff --git a/requirements/requirements.txt b/requirements/requirements.txt
index ef2c016..756a65d 100644
--- a/requirements/requirements.txt
+++ b/requirements/requirements.txt
@@ -2,7 +2,7 @@
 # This file is autogenerated by pip-compile with Python 3.11
 # by the following command:
 #
-#    pip-compile
+#    pip-compile requirements.in
 #
 aiohttp==3.8.4
     # via
@@ -15,7 +15,7 @@ async-timeout==4.0.2
 attrs==23.1.0
     # via aiohttp
 beautifulsoup4==4.12.2
-    # via -r requirements/requirements.in
+    # via -r requirements.in
 certifi==2023.5.7
     # via requests
 charset-normalizer==3.1.0
@@ -23,9 +23,11 @@ charset-normalizer==3.1.0
     #   aiohttp
     #   requests
 click==8.1.4
-    # via -r requirements/requirements.in
+    # via -r requirements.in
 dataclasses-json==0.5.8
     # via langchain
+faiss-cpu==1.7.4
+    # via -r requirements.in
 frozenlist==1.3.3
     # via
     #   aiohttp
@@ -33,19 +35,17 @@ frozenlist==1.3.3
 gitdb==4.0.10
     # via gitpython
 gitpython==3.1.32
-    # via -r requirements/requirements.in
-greenlet==2.0.2
-    # via sqlalchemy
+    # via -r requirements.in
 idna==3.4
     # via
     #   requests
     #   yarl
 langchain==0.0.231
-    # via -r requirements/requirements.in
+    # via -r requirements.in
 langchainplus-sdk==0.0.20
     # via langchain
 loguru==0.7.0
-    # via -r requirements/requirements.in
+    # via -r requirements.in
 markdown-it-py==3.0.0
     # via rich
 marshmallow==3.19.0
@@ -69,13 +69,13 @@ numpy==1.25.0
     #   langchain
     #   numexpr
 openai==0.27.8
-    # via -r requirements/requirements.in
+    # via -r requirements.in
 openapi-schema-pydantic==1.2.4
     # via langchain
 packaging==23.1
     # via marshmallow
 prompt-toolkit==3.0.39
-    # via -r requirements/requirements.in
+    # via -r requirements.in
 pydantic==1.10.9
     # via
     #   langchain
@@ -85,7 +85,7 @@ pygments==2.15.1
     # via rich
 pyyaml==6.0
     # via
-    #   -r requirements/requirements.in
+    #   -r requirements.in
     #   langchain
 regex==2023.6.3
     # via tiktoken
@@ -96,7 +96,7 @@ requests==2.31.0
     #   openai
     #   tiktoken
 rich==13.4.2
-    # via -r requirements/requirements.in
+    # via -r requirements.in
 smmap==5.0.0
     # via gitdb
 soupsieve==2.4.1
@@ -108,7 +108,7 @@ tenacity==8.2.2
     #   langchain
     #   langchainplus-sdk
 tiktoken==0.4.0
-    # via -r requirements/requirements.in
+    # via -r requirements.in
 tqdm==4.65.0
     # via openai
 typing-extensions==4.6.3
diff --git a/tests/test_coder.py b/tests/test_coder.py
index d5221eb..f0dff81 100644
--- a/tests/test_coder.py
+++ b/tests/test_coder.py
@@ -1,6 +1,6 @@
 from aicodebot.coder import Coder
 from aicodebot.helpers import create_and_write_file
-import os
+import os, pytest
 
 
 def test_generate_directory_structure(tmp_path):
@@ -113,3 +113,18 @@ def test_git_diff_context(temp_git_repo):
     commit = temp_git_repo.head.commit.hexsha
     diff = Coder.git_diff_context(commit)
     assert "renamedfile.txt" in diff
+
+    def test_parse_github_url():
+        # Test with https URL
+        owner, repo = Coder.parse_github_url("https://github.com/owner/repo.git")
+        assert owner == "owner"
+        assert repo == "repo"
+
+        # Test with git URL
+        owner, repo = Coder.parse_github_url("git@github.com:owner/repo.git")
+        assert owner == "owner"
+        assert repo == "repo"
+
+        # Test with invalid URL
+        with pytest.raises(ValueError):
+            Coder.parse_github_url("not a valid url")