diff --git a/.env.example b/.env.example
index cd6c51f..67cb62c 100644
--- a/.env.example
+++ b/.env.example
@@ -18,6 +18,10 @@ ANTHROPIC_MODEL_CHAT=claude-3-5-sonnet-latest
 # --- Google Gemini (optional, used by MCP server) ---
 GOOGLE_API_KEY=
 
+# --- Search Provider ---
+SEARCH_PROVIDER=duckduckgo       # duckduckgo | tavily
+TAVILY_API_KEY=                  # Required when SEARCH_PROVIDER=tavily
+
 # --- Storage ---
 CHROMA_DIR=.chroma
 SQLITE_PATH=.sqlite/agent.db
diff --git a/requirements.txt b/requirements.txt
index d10d119..2c0f0c6 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -17,6 +17,7 @@ typing-extensions>=4.12.2
 
 # Discovery
 duckduckgo-search>=6.2.10
+tavily-python>=0.3.0
 httpx>=0.27.0
 trafilatura>=1.10.0
 beautifulsoup4>=4.12.3
diff --git a/src/agentic_ai/layers/tools.py b/src/agentic_ai/layers/tools.py
index 4818ed8..53d3d2a 100644
--- a/src/agentic_ai/layers/tools.py
+++ b/src/agentic_ai/layers/tools.py
@@ -6,8 +6,8 @@
 
 from ..tools.knowledge import KbAdd, KbSearch
 from ..tools.ops import Calculator, Emailer, FileWrite
-from ..tools.webtools import WebFetch, WebSearch
+from ..tools.webtools import WebFetch, get_web_search_tool
 
 
 def registry() -> List[BaseTool]:
-    return [WebSearch(), WebFetch(), KbSearch(), KbAdd(), Calculator(), FileWrite(), Emailer()]
+    return [get_web_search_tool(), WebFetch(), KbSearch(), KbAdd(), Calculator(), FileWrite(), Emailer()]
diff --git a/src/agentic_ai/tools/webtools.py b/src/agentic_ai/tools/webtools.py
index 5561259..ca847e0 100644
--- a/src/agentic_ai/tools/webtools.py
+++ b/src/agentic_ai/tools/webtools.py
@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import json
+import os
 
 import httpx
 import trafilatura
@@ -23,6 +24,38 @@ def _run(self, query: str) -> str:
         return json.dumps(list(results), ensure_ascii=False)
 
 
+class TavilyWebSearch(BaseTool):
+    name: str = "web_search"
+    description: str = (
+        "Search the web. Input: a natural language query. "
+        "Output: JSON list of {title, url, snippet}."
+    )
+
+    @retry(stop=stop_after_attempt(3), wait=wait_exponential(multiplier=1, min=1, max=4))
+    def _run(self, query: str) -> str:
+        from tavily import TavilyClient
+
+        client = TavilyClient()
+        response = client.search(query=query, max_results=8)
+        results = [
+            {
+                "title": r.get("title", ""),
+                "url": r.get("url", ""),
+                "snippet": r.get("content", ""),
+            }
+            for r in response.get("results", [])
+        ]
+        return json.dumps(results, ensure_ascii=False)
+
+
+def get_web_search_tool() -> BaseTool:
+    """Return the active web search tool based on SEARCH_PROVIDER env var."""
+    provider = os.environ.get("SEARCH_PROVIDER", "duckduckgo").lower()
+    if provider == "tavily":
+        return TavilyWebSearch()
+    return WebSearch()
+
+
 class WebFetch(BaseTool):
     name: str = "web_fetch"
     description: str = "Fetch a URL and extract main readable text. Input: URL. Output: text."