How to use prebuilt tools

AgentFlow ships a set of production-ready tools in agentflow.prebuilt.tools. Drop them into any ToolNode or pass them directly to a prebuilt agent's tools list.

from agentflow.prebuilt.tools import (
    fetch_url,
    file_read,
    file_write,
    file_search,
    safe_calculator,
    google_web_search,
    vertex_ai_search,
    memory_tool,
    make_user_memory_tool,
    make_agent_memory_tool,
    create_handoff_tool,
)

---

fetch_url

Fetches the text content of any public HTTP/HTTPS URL. Blocks private/loopback IP addresses, enforces a configurable timeout, and truncates long responses.

from agentflow.core.graph import Agent, ToolNode
from agentflow.prebuilt.tools import fetch_url

tool_node = ToolNode([fetch_url])

agent = Agent(
    model="gpt-4o",
    tool_node=tool_node,
    system_prompt=[{"role": "system", "content": "You are a research assistant."}],
)

Tool schema (what the LLM sees):

Parameter	Type	Default	Description
url	str	required	Public HTTP/HTTPS URL to fetch.
timeout	float	10.0	Request timeout in seconds (max 30).
max_chars	int	20000	Maximum characters to return.

The tool returns a JSON string:

{
  "url": "https://example.com",
  "status_code": 200,
  "content_type": "text/html",
  "content": "...",
  "truncated": false
}

Tags: ["web", "fetch", "network"]

---

safe_calculator

Evaluates arithmetic expressions without exposing __builtins__. Safe for production.

from agentflow.prebuilt.tools import safe_calculator

tool_node = ToolNode([safe_calculator])

Tool schema:

Parameter	Type	Description
expression	str	Arithmetic expression to evaluate, e.g. "123 * 456 + 789".

Returns the result as a string, or an error message if evaluation fails.

---

file_read, file_write, file_search

Local filesystem tools. All three enforce that the path is within the current working directory or an explicit allowed root.

from agentflow.prebuilt.tools import file_read, file_write, file_search

tool_node = ToolNode([file_read, file_write, file_search])

file_read

Parameter	Type	Description
path	str	Relative or absolute path to the file.
encoding	str	File encoding (default "utf-8").

Returns file content as a string.

file_write

Parameter	Type	Description
path	str	Path to write to.
content	str	Text content to write.
mode	str	"w" (overwrite, default) or "a" (append).

Returns a success or error message.

file_search

Parameter	Type	Description
pattern	str	Glob pattern, e.g. "*.py" or "**/*.md".
root	str	Root directory to search from (default: current working directory).

Returns a JSON list of matching file paths.

---

google_web_search

Calls the Google Custom Search API. Requires GOOGLE_API_KEY and GOOGLE_CSE_ID environment variables.

export GOOGLE_API_KEY=your-google-api-key
export GOOGLE_CSE_ID=your-custom-search-engine-id

from agentflow.prebuilt.tools import google_web_search

tool_node = ToolNode([google_web_search])

Tool schema:

Parameter	Type	Default	Description
query	str	required	Search query string.
num_results	int	5	Number of results to return (max 10).

Returns a JSON list of {"title": ..., "url": ..., "snippet": ...} objects.

---

vertex_ai_search

Calls Google Vertex AI Search. Requires Google Cloud credentials and a Vertex AI data store ID.

from agentflow.prebuilt.tools import vertex_ai_search

tool_node = ToolNode([vertex_ai_search])

Configure via environment variables:

export GOOGLE_CLOUD_PROJECT=your-gcp-project
export VERTEX_AI_DATA_STORE_ID=your-data-store-id

---

Memory tools

Memory tools let the LLM search and write long-term user or agent memories. They are designed to be used with MemoryConfig.

memory_tool

A general-purpose memory search and write tool for use without MemoryConfig.

from agentflow.prebuilt.tools import memory_tool
from agentflow.storage.store import create_local_qdrant_store, OpenAIEmbedding

store = create_local_qdrant_store("./qdrant_data", OpenAIEmbedding())
tool = memory_tool(store)

tool_node = ToolNode([tool])

make_user_memory_tool and make_agent_memory_tool

These are used internally by MemoryConfig; you can also call them directly if you need more control.

from agentflow.prebuilt.tools import make_user_memory_tool, make_agent_memory_tool
from agentflow.storage.store import MemoryConfig, UserMemoryConfig

config = MemoryConfig(store=store)
user_tool = make_user_memory_tool(config)
agent_tool = make_agent_memory_tool(config)

tool_node = ToolNode([user_tool, agent_tool])

The typical pattern is to let Agent(..., memory=MemoryConfig(...)) inject these tools automatically rather than registering them manually.

---

create_handoff_tool

Creates a handoff tool that transfers control from one agent to another in multi-agent graphs (swarm or supervisor patterns). See how-to/python/handoff-between-agents for the full handoff guide.

from agentflow.prebuilt.tools import create_handoff_tool

handoff_to_billing = create_handoff_tool(
    agent_name="billing",
    description="Transfer the user to the billing agent for payment questions.",
)

tool_node = ToolNode([handoff_to_billing])

Parameters:

Parameter	Type	Description
agent_name	str	Name of the target agent node in the graph.
description	str	Description shown to the LLM to help it decide when to hand off.

---

Composing tools

All prebuilt tools can be mixed with custom tools in a single ToolNode:

from agentflow.core.graph import Agent, StateGraph, ToolNode
from agentflow.prebuilt.tools import fetch_url, safe_calculator
from agentflow.utils.decorators import tool

@tool(name="get_exchange_rate", tags=["finance"])
async def get_exchange_rate(from_currency: str, to_currency: str) -> str:
    """Get the current exchange rate between two currencies."""
    return f"1 {from_currency} = 1.08 {to_currency}"

tool_node = ToolNode([fetch_url, safe_calculator, get_exchange_rate])

agent = Agent(
    model="gpt-4o",
    tool_node=tool_node,
)

---

Prebuilt tool tags reference

Tool	Tags
fetch_url	["web", "fetch", "network"]
safe_calculator	["math", "calculator"]
file_read	["file", "read"]
file_write	["file", "write"]
file_search	["file", "search"]
google_web_search	["search", "web", "google"]
vertex_ai_search	["search", "vertex", "google"]

Use Agent(..., tools_tags={"search"}) to expose only search-tagged tools to a particular agent.

---

What you learned

• fetch_url fetches public URLs safely with timeout and size limits.
• safe_calculator evaluates arithmetic without exposing Python builtins.
• file_read, file_write, file_search provide controlled filesystem access.
• google_web_search and vertex_ai_search enable live web search capabilities.
• Memory tools are best used via Agent(..., memory=MemoryConfig(...)) rather than added manually.
• create_handoff_tool enables agent-to-agent handoffs in multi-agent workflows.

Next steps

• Use prebuilt agents for ready-made agent wrappers.
• Use the @tool decorator to build your own tools.