diff --git a/tools.py b/tools.py
new file mode 100644
index 0000000..d2fe16a
--- /dev/null
+++ b/tools.py
@@ -0,0 +1,56 @@
+"""Tool definitions for the FAQ bot.
+
+The module defines two tools:
+1. `search_course_docs` – a semantic search over the local Chroma FAQ store.
+2. `fetch_course_meta` – a simple HTTP GET to a local JSON file that mimics an MCP
+   metadata service.
+
+Both tools are decorated with `@tool` so that LangChain can expose them to the
+agent.
+"""
+
+import json
+import os
+from pathlib import Path
+from typing import Any
+
+import httpx
+from langchain.tools import tool
+
+from config import COURSE_META_JSON
+from vector_store import search_course_docs
+
+# ---------------------------------------------------------------------------
+# 1. Semantic search tool
+# ---------------------------------------------------------------------------
+@tool("search_course_docs", "Search local FAQ documents.")
+async def search_course_docs_tool(query: str, k: int = 3) -> str:
+    """Return top‑k relevant FAQ snippets for *query*.
+
+    The function is asynchronous to match the signature expected by LangChain
+    agents. It simply forwards the call to the synchronous helper in
+    ``vector_store``.
+    """
+    return search_course_docs(query, k)
+
+# ---------------------------------------------------------------------------
+# 2. Metadata fetch tool – MCP‑style
+# ---------------------------------------------------------------------------
+@tool("fetch_course_meta", "Get course metadata from a local JSON file.")
+async def fetch_course_meta_tool(query: str) -> str:
+    """Return JSON string of metadata matching *query*.
+
+    The function reads a static JSON file. In a real deployment this would be a
+    HTTP request to an MCP server. For the purposes of the assignment we keep
+    it simple and local.
+    """
+    if not Path(COURSE_META_JSON).exists():
+        return f"Metadata file {COURSE_META_JSON} not found."
+    data = json.loads(Path(COURSE_META_JSON).read_text(encoding="utf-8"))
+    # Very naive lookup: return the whole file if query is empty or substring
+    if not query or query.lower() in json.dumps(data).lower():
+        return json.dumps(data, indent=2)
+    return f"No metadata found for query: {query}."
+
+# Expose the tools for external imports
+__all__ = ["search_course_docs_tool", "fetch_course_meta_tool"]
\ No newline at end of file