Skip to main content
Deep Agents can call any tool you define, any LangChain tool, and tools from any MCP server. Pass them to create_deep_agent via the tools= parameter alongside the built-in harness tools for planning, file management, and subagent spawning. 
from deepagents import create_deep_agent

agent = create_deep_agent(
    model="anthropic:claude-sonnet-4-6",
    tools=[search, fetch_url, run_query],
)

​
Custom tools

Pass any callable — plain functions, LangChain @tool-decorated functions, or tool dicts — directly to tools=. Deep Agents infers the tool schema from the function signature and docstring, so you don’t need to define a separate schema in most cases. 
import os
from typing import Literal
from tavily import TavilyClient
from deepagents import create_deep_agent

tavily_client = TavilyClient(api_key=os.environ["TAVILY_API_KEY"])


def internet_search(
    query: str,
    max_results: int = 5,
    topic: Literal["general", "news", "finance"] = "general",
    include_raw_content: bool = False,
):
    """Run a web search"""
    return tavily_client.search(
        query,
        max_results=max_results,
        include_raw_content=include_raw_content,
        topic=topic,
    )


agent = create_deep_agent(
    model="google_genai:gemini-3.5-flash",
    tools=[internet_search],
)
For full details on defining and using LangChain tools (tool dicts, StructuredTool, return types, error handling, and more), see Tools.

​
MCP tools

Deep Agents fully support Model Context Protocol (MCP) — the open standard for connecting agents to external services. Load tools from any MCP server and pass them directly to create_deep_agent.
MCP is an open protocol that lets agents connect to a growing ecosystem of servers — databases, APIs, file systems, browsers, and more — through a standard interface. Instead of writing custom integration code for each service, you point Deep Agents at an MCP server and it gets all the tools that server exposes. Install langchain-mcp-adapters to connect to MCP servers: 
pip install langchain-mcp-adapters

import asyncio
from langchain_mcp_adapters.client import MultiServerMCPClient
from deepagents import create_deep_agent

async def main():
    async with MultiServerMCPClient(
        {
            "my_server": {
                "transport": "http",
                "url": "http://localhost:8000/mcp",
            }
        }
    ) as client:
        tools = await client.get_tools()

        agent = create_deep_agent(
            model="openai:gpt-5.5",
            tools=tools,
        )

        result = await agent.ainvoke(
            {"messages": [{"role": "user", "content": "Use the MCP server to help me."}]},
            config={"configurable": {"thread_id": "1"}},
        )

asyncio.run(main())
For detailed configuration options — including stdio servers, OAuth authentication, tool filtering, and stateful sessions — see the full MCP guide.

​
Built-in harness tools

In addition to the tools you provide, every Deep Agent comes with a built-in set of tools from the harness:
ToolDescription
lsList files in a directory
read_fileRead file contents (with pagination and multimodal support)
write_fileCreate new files
edit_filePerform exact string replacements in files
globFind files matching a glob pattern
grepSearch file contents
executeRun shell commands (sandbox backends only)
taskSpawn a subagent to handle a delegated task
write_todosManage a structured todo list
For a full breakdown of what each built-in tool does, see Harness overview.

​
Multimodal tool outputs

Custom tools can return plain text or standard content blocks (text, images, audio, video, and files) when the selected model supports multimodal tool results. The built-in read_file tool also returns multimodal blocks for supported non-text file types. Return a string for text-only results, or an ordered list of content blocks for text plus media or interleaved multimodal output. See Multimodal and Tool return values for examples and context-compression considerations.
Connect these docs to Claude, VSCode, and more via MCP for real-time answers.
Edit this page on GitHub or file an issue.