AWS Bedrock AgentCore search tools

Add Valyu search tools to a Strands agent, then choose local execution or an AgentCore deployment. Gateway authentication, permissions and audit logging depend on how you configure AWS. Start with a local query before provisioning infrastructure.

Available search tools

Tool	Best For	Data Sources
`webSearch`	News, current events, general information	Web pages, news sites
`financeSearch`	Stock prices, earnings, market analysis	Stocks, forex, crypto, balance sheets
`paperSearch`	Literature review, academic research	arXiv, PubMed, bioRxiv, medRxiv
`bioSearch`	Medical research, drug information	PubMed, clinical trials, FDA labels
`patentSearch`	Prior art, IP research	USPTO patents
`secSearch`	Company analysis, due diligence	SEC filings (10-K, 10-Q, 8-K), insider transactions (Form 4)
`economicsSearch`	Economic indicators, policy research	BLS, FRED, World Bank, US Spending

Quick start

Installation

# For local development with Strands Agents
pip install "valyu-agentcore[strands]"

# For AWS AgentCore Gateway/Runtime deployment
pip install "valyu-agentcore[agentcore]"

Environment Setup

export VALYU_API_KEY="your-valyu-api-key"
export AWS_REGION="us-east-1"  # Optional, defaults to us-east-1

Get your free API key at platform.valyu.ai - includes $10 free credits ($20 with a company email), no credit card required.

Your First Agent

Create an AI agent with web search capability in just a few lines:

from valyu_agentcore import webSearch
from strands import Agent
from strands.models import BedrockModel

# Create an agent with Valyu search
agent = Agent(
    model=BedrockModel(model_id="us.anthropic.claude-sonnet-4-20250514-v1:0"),
    tools=[webSearch()],
)

# Ask anything
response = agent("What are the latest developments in quantum computing?")
print(response)

Multi-Tool Agent

Combine multiple search tools for comprehensive research:

from valyu_agentcore import webSearch, financeSearch, secSearch, paperSearch
from strands import Agent
from strands.models import BedrockModel

# Financial research agent with multiple data sources
agent = Agent(
    model=BedrockModel(model_id="us.anthropic.claude-sonnet-4-20250514-v1:0"),
    tools=[
        webSearch(),      # Current news and web content
        financeSearch(),  # Stock prices and financial data
        secSearch(),      # SEC filings and disclosures
        paperSearch(),    # Academic research
    ],
)

response = agent("Analyze NVIDIA's competitive position in the AI chip market")
print(response)

Tool groups

The ValyuTools wrapper provides ready-made groupings - tools.all(), tools.financial_tools() (finance, SEC, economics), and tools.research_tools() (paper, bio, patent):

from valyu_agentcore import ValyuTools
from strands import Agent
from strands.models import BedrockModel

tools = ValyuTools(max_num_results=5)

agent = Agent(
    model=BedrockModel(model_id="us.anthropic.claude-sonnet-4-20250514-v1:0"),
    tools=tools.financial_tools(),
)

Deployment options

Option	When to use	Trade-off
Local	Prototyping and testing. Direct API calls.	No centralized key management or audit logging
AgentCore Gateway	Production. OAuth, centralized keys, CloudTrail.	Requires AWS infrastructure setup
AgentCore Runtime	Fully managed, auto-scaling deployment.	More setup, AWS costs

Gateway setup

from valyu_agentcore.gateway import setup_valyu_gateway, GatewayAgent

config = setup_valyu_gateway()  # creates Gateway + Cognito resources
with GatewayAgent.from_config() as agent:
    print(agent("Search for NVIDIA SEC filings"))

Add Valyu to an existing gateway with add_valyu_target(gateway_id="...", region="us-east-1").

Runtime deployment

cd examples/runtime
agentcore configure --entrypoint agent.py --non-interactive --name valyuagent
agentcore launch
agentcore invoke '{"prompt": "What is NVIDIA stock price?"}'

CloudFormation

Deploy the full stack (IAM role, Cognito OAuth pool, Secrets Manager secret, CloudWatch logs) in one command:

aws cloudformation create-stack \
  --stack-name valyu-agentcore \
  --template-body file://cloudformation/valyu-gateway.yaml \
  --parameters \
    ParameterKey=ValyuApiKey,ParameterValue=your-api-key \
    ParameterKey=GatewayName,ParameterValue=valyu-search-gateway \
  --capabilities CAPABILITY_IAM

Tool configuration

from valyu_agentcore import financeSearch

tool = financeSearch(
    api_key="val_xxx",        # or VALYU_API_KEY env var
    search_type="all",        # "all", "web", "proprietary", "news"
    max_num_results=10,
    relevance_threshold=0.7,  # 0-1 quality filter
    included_sources=["sec.gov"],
    category="quarterly earnings",
)

Example: financial analyst agent

Use a system prompt to route requests between tools. This example demonstrates retrieval and synthesis; evaluate outputs and configure operational controls before using it in a production workflow:

from valyu_agentcore import financeSearch, secSearch, webSearch
from strands import Agent
from strands.models import BedrockModel

agent = Agent(
    model=BedrockModel(model_id="us.anthropic.claude-sonnet-4-20250514-v1:0"),
    system_prompt="""You are a senior financial analyst. Use financeSearch for prices
    and earnings, secSearch for filings and risk factors, webSearch for recent news.
    Be specific with numbers and cite sources.""",
    tools=[financeSearch(), secSearch(), webSearch()],
)

response = agent("Analyze Apple's most recent quarterly earnings and outlook")

Security best practices

Keep API keys in VALYU_API_KEY or AWS Secrets Manager - never hardcode.
Configure CloudTrail for supported AWS events and application logs for tool activity.
Use AgentCore Gateway with Cognito OAuth in production.
Apply least-privilege IAM (invoke-only for applications).
Rotate Valyu API keys and Cognito client secrets regularly.

Troubleshooting

No module named 'strands' - install with pip install "valyu-agentcore[strands]".
Gateway sync slow - targets take ~2 minutes; add_valyu_target() waits automatically.
OAuth token errors - check the Cognito client scope (gateway-name/invoke) and that client_credentials is enabled.
Bedrock model access - request model access in the AWS Bedrock console.

API reference

Tools: webSearch, financeSearch, paperSearch, bioSearch, patentSearch, secSearch, economicsSearch.ValyuTools: .all(), .financial_tools() (finance, SEC, economics), .research_tools() (paper, bio, patent).Gateway: setup_valyu_gateway(), add_valyu_target(), cleanup_valyu_gateway(), GatewayAgent.from_config().

Additional resources

GitHub Repository

Source code, examples, and CloudFormation templates