A new class of research is emerging where giving models tools for structured thinking, like a scratchpad, greatly improves their reasoning capabilities. For example, by giving a model a “think” tool, we can greatly improve its reasoning capabilities by providing a dedicated space for working through the problem. This is a simple, yet effective approach to add reasoning to non-reasoning models. First published by Anthropic in this blog post, this technique has been practiced by many AI Engineers (including our own team) long before it was published.

​
v0: The Think Tool

The first version of the Think Tool was published by Anthropic in this blog post.
claude_thinking_tools.py
from agno.agent import Agent
from agno.models.anthropic import Claude
from agno.tools.thinking import ThinkingTools
from agno.tools.duckduckgo import DuckDuckGoTools

reasoning_agent = Agent(
    model=Claude(id="claude-3-7-sonnet-latest"),
    tools=[
        ThinkingTools(add_instructions=True),
        DuckDuckGoTools(
            search=True,
            news=True,
        ),
    ],
    instructions="Use tables where possible",
    markdown=True,
)

if __name__ == "__main__":
    reasoning_agent.print_response(
        "Write a report on the latest developments in AI technology. Only the report, no other text.",
        stream=True,
        show_full_reasoning=True,
        stream_intermediate_steps=True,
    )

​
v1: The Reasoning Tools

While the v0 Think Tool is a great start, it is limited in that it only allows for a thinking space. The v1 Reasoning Tools take this one step further by allowing the Agent to analyze the results of their actions (i.e. tool calls), greatly improving the Agents’ ability to solve problems that require sequential tool calls. ReasoningTools = think + analyze
claude_reasoning_tools.py
from agno.agent import Agent
from agno.models.anthropic import Claude
from agno.tools.reasoning import ReasoningTools
from agno.tools.duckduckgo import DuckDuckGoTools

reasoning_agent = Agent(
    model=Claude(id="claude-3-7-sonnet-20250219"),
    tools=[
        ReasoningTools(add_instructions=True),
        DuckDuckGoTools(search=True, news=True),
    ],
    show_tool_calls=True,
)
reasoning_agent.print_response(
    "Write a report comparing recent developments in renewable energy vs traditional energy", stream=True, markdown=True
)

​
v2: The Knowledge Tools

The Knowledge Tools take the v1 Reasoning Tools one step further by allowing the Agent to search a knowledge base and analyze the results of their actions. KnowledgeTools = think + search + analyze
knowledge_tools.py
import os
from agno.agent import Agent
from agno.embedder.openai import OpenAIEmbedder
from agno.knowledge.url import UrlKnowledge
from agno.models.openai import OpenAIChat
from agno.tools.knowledge import KnowledgeTools
from agno.vectordb.lancedb import LanceDb, SearchType


agno_docs = UrlKnowledge(
    urls=["https://docs-v1.agno.com/llms-full.txt"],

    vector_db=LanceDb(
        uri="tmp/lancedb",
        table_name="agno_docs",
        search_type=SearchType.hybrid,
        embedder=OpenAIEmbedder(id="text-embedding-3-small"),
    ),
)


knowledge_tools = KnowledgeTools(
    knowledge=agno_docs,
    think=True,   
    search=True,  
    analyze=True,  
    add_few_shot=True, 
)


agent = Agent(
    model=OpenAIChat(id="gpt-4o"),
    tools=[knowledge_tools],
    show_tool_calls=True, 
    markdown=True, 
)


agno_docs.load(recreate=True)


agent.print_response("How do I build multi-agent teams with Agno?", stream=True)

​
Developer Resources