strands-agents/tools

Strands Agents Tools

What it is – A Python package that bundles a large collection of ready‑to‑use tools (functions) that an LLM‑driven agent can call. The tools cover everything from simple file I/O to web search, AWS service calls, image/video generation, desktop automation, and multi‑agent orchestration. They are meant to be plugged into the Strands Agents SDK so that a language model can invoke them with a single line of code.

Key capabilities

Category	Example tools	Typical use case
File & editor	file_read, file_write, editor (deprecated)	Load config files, write results, edit code snippets with syntax highlighting
Shell & OS	shell (deprecated), environment (deprecated)	Run system commands, query env vars (use with caution on Windows)
Web & search	tavily_search, tavily_extract, exa_search, exa_get_contents	Real‑time web search, content extraction, site crawling
HTTP / API	http_request	Call any REST endpoint with auth support
Messaging	slack (deprecated)	Post or read messages in a Slack workspace
Python execution	python_repl, code_interpreter	Execute arbitrary Python (or JS/TS via the sandboxed interpreter) with state persistence
Math & reasoning	calculator (deprecated), think (deprecated)	Symbolic calculations, multi‑step reasoning (now handled natively by the SDK)
AWS integration	use_aws, nova_reels, agent_core_memory	Interact with S3, Bedrock, other AWS services
Media generation	generate_image, generate_image_stability, search_video, chat_video	Create images, search and analyse videos
Memory back‑ends	mem0_memory, mongodb_memory, elasticsearch_memory	Store and retrieve user/agent memories across runs
Browser automation	browser (Chromium)	Scrape pages, fill forms, run UI tests
Desktop automation	use_computer	Move mouse, click, type, take screenshots
Multi‑agent orchestration	swarm, agent_graph, graph, use_agent, use_llm	Run several agents in parallel, build deterministic DAG pipelines, nest agents with model switching
Scheduling & control	cron (deprecated), sleep (deprecated), stop, handoff_to_user	Schedule recurring jobs, pause execution, gracefully stop, ask the user for confirmation
Miscellaneous	journal, rss (deprecated), bright_data, speak	Structured logging, RSS feed handling, web‑scraping service, text‑to‑speech

Installation

# Basic install (core tools only)
pip install strands-agents-tools

# Install optional extras you need, e.g. memory back‑ends, browser, RSS, computer automation
pip install "strands-agents-tools[mem0_memory,use_browser,rss,use_computer]"

For development, clone the repo, create a virtual environment, and run pip install -e ".[dev]".

How to use – The README shows the typical pattern:

from strands_tools import A2AClientToolProvider, Agent

provider = A2AClientToolProvider(known_agent_urls=["http://localhost:9000"])
agent = Agent(tools=provider.tools)

# Call a tool directly on the agent
result = agent.tool.file_read(path="config.yaml")

Each tool is exposed as a method on agent.tool. Arguments are passed as keyword parameters, and the return value is whatever the underlying operation produces (e.g., file contents, JSON response, image bytes, etc.).

Safety notice – Many tools execute code, touch the filesystem, or call external services. The package marks such tools as experimental and adds deprecation warnings for older wrappers. Production deployments should perform an independent security review and follow the project’s Responsible AI guidance.

Roadmap / status – The repository is community‑driven and actively maintained (see the commit‑activity badge). Several tools are already deprecated in favor of native SDK features; the maintainers plan to eventually archive the repo once most functionality lives directly in the Strands SDK.

---

TL;DR – Strands Agents Tools is a Python toolbox that lets LLM‑powered agents perform real‑world actions (file I/O, web search, AWS calls, desktop automation, multi‑agent pipelines, etc.) with a few lines of code. Install via pip, plug the tools into a Strands Agent, and call them as methods. Be aware of security implications and the ongoing deprecation of older wrappers.