ScriptsApr 7, 2026·2 min read

Toolhouse — Managed AI Tool Infrastructure

Cloud-hosted tool execution for AI agents. Toolhouse provides 100+ pre-built tools (web search, code exec, email) with managed auth, logging, and one SDK call.

AG
Agent Toolkit · Community
Quick Use

Use it first, then decide how deep to go

This block should tell both the user and the agent what to copy, install, and apply first.

pip install toolhouse
from toolhouse import Toolhouse
from anthropic import Anthropic

th = Toolhouse()
client = Anthropic()

messages = [{"role": "user", "content": "Search the web for Claude Code tips"}]

response = client.messages.create(
    model="claude-sonnet-4-20250514",
    messages=messages,
    tools=th.get_tools(),
)

# Execute tool calls
tool_results = th.run_tools(response)

What is Toolhouse?

Toolhouse provides managed tool infrastructure for AI agents. Instead of building and hosting tools yourself, Toolhouse offers 100+ pre-built, cloud-hosted tools — web search, code execution, email, file operations — accessible via a single SDK call with managed authentication and logging.

Answer-Ready: Toolhouse is a managed AI tool infrastructure providing 100+ pre-built cloud-hosted tools (web search, code execution, email) accessible via one SDK call. Handles auth, execution, and logging so you focus on agent logic.

Best for: AI agent developers who want tools without infrastructure overhead. Works with: Anthropic, OpenAI, LangChain, Vercel AI SDK. Setup time: Under 2 minutes.

Core Features

1. Pre-Built Tool Categories

Category Tools
Web Search, scrape, read URL, screenshot
Code Execute Python, Node.js, shell
Communication Send email, Slack message
Data Query databases, parse files
AI Generate images, transcribe audio
Storage Read/write files, manage S3

2. One-Line Integration

# Anthropic
tools = th.get_tools()  # Returns tool definitions
results = th.run_tools(response)  # Executes tool calls

# OpenAI
from openai import OpenAI
client = OpenAI()
tools = th.get_tools(provider="openai")

3. Tool Bundles

Group tools by use case:

# Only web tools
tools = th.get_tools(bundle="web")

# Only code execution
tools = th.get_tools(bundle="code")

# Custom selection
tools = th.get_tools(tools=["web_search", "send_email", "execute_python"])

4. Managed Execution

Tools run in Toolhouse's cloud:

  • Sandboxed code execution
  • Managed API keys (no credential sharing with LLMs)
  • Automatic retries and error handling
  • Request logging and analytics

5. Custom Tools

@th.register_tool("my_crm_lookup")
def crm_lookup(company_name: str) -> dict:
    # Your custom logic
    return {"revenue": "10M", "employees": 50}

Architecture

Your Agent → Anthropic/OpenAI API → Tool call response
                                         ↓
                                    Toolhouse SDK
                                         ↓
                                  Toolhouse Cloud
                                  (executes tools)
                                         ↓
                                   Tool results
                                         ↓
                                  Back to your agent

FAQ

Q: Is Toolhouse free? A: Free tier with 1,000 tool executions/month. Paid plans for higher volume.

Q: Does my data pass through Toolhouse? A: Tool inputs/outputs pass through Toolhouse for execution. Code runs in sandboxed environments.

Q: Can I self-host? A: Currently cloud-only. Self-hosting is on the roadmap.

🙏

Source & Thanks

Created by Toolhouse. SDK is open-source.

toolhouseai/toolhouse-sdk-python Docs: docs.toolhouse.ai

Discussion

Sign in to join the discussion.
No comments yet. Be the first to share your thoughts.

Related Assets