|

||

THE NPM FOR AI TOOLS

232PACKAGES

/

797TOOLS

/

194.8KDOWNLOADS

Discover and use npm packages as AI agent tools. Works with Claude, GPT, LangChain, and any MCP-compatible client.

Try: “web scraper”, “discord”, “code interpreter”

get started in 60 seconds

one command. every editor.

Configure your AI tools across Claude Code, Cursor, Windsurf, and more with a single install command. Choose your collections, pick your editors, run the script.

1

pick your collections

Select which tool collections you want in your editors. Your collections, or any public collection.

2

choose your editors

Map each collection to Claude Code, Cursor, Windsurf, or Claude Desktop. Different collections for different editors.

3

run one command

Paste in your terminal. The script auto-configures MCP servers in every selected editor. Done.

install script

$

curl -fsSL https://tpmjs.com/install.sh | bash

tip

visit tpmjs.com/setup first to select your collections and get a personalized command with your setup token included.

>_Claude Code

CClaude Desktop

{;}Cursor

WWindsurf

0

Published Tools

Auto-synced from npm

0

Active Developers

Last 7 days

0

Total Executions

All-time simulations

0.0s

Avg Response

Execution latency

Recent Activity

↻calendar-sync updated12s ago

↻slack-bot updated16s ago

↻sql-query updated18s ago

↻github-manager updated1s ago

+image-processor published5s ago

platform capabilities

everything you need

From discovery to execution, TPMJS provides the complete infrastructure for AI tool development and deployment.

tool registry

Browse 797 AI tools from npm. Auto-discovered within minutes of publication with quality scoring and health monitoring.

omega agent

Chat with an AI that dynamically discovers and executes tools based on your requests. No configuration needed.

collections

Curate tool sets for specific use cases. Add test scenarios to validate behavior and generate living documentation.

custom agents

Build AI agents with your choice of LLM, custom prompts, and curated tool collections. Share publicly or keep private.

mcp protocol

Works with Claude Desktop, Cursor, Windsurf, and any MCP-compatible client. One URL, instant access to all tools.

secure execution

sandboxed

Every tool runs in an isolated sandbox with rate limiting and timeout handling. Your credentials are encrypted at rest.

test scenarios

AI-generated test scenarios validate tool behavior. Track pass rates, execution times, and quality scores.

living skills

Documentation that evolves from real usage. Skills emerge from question patterns and proven behaviors.

developer sdk

Publish tools with one keyword. Full TypeScript support, Vercel AI SDK integration, and automatic schema extraction.

Featured Tools

Production-ready tools you can use today. Add to your AI agent in one line.

@thomasdavis/unsandbox(executeCodeAsync)

Execute code asynchronously in a secure sandbox. Returns a job_id immediately. Use getJob to check status and retrieve results. Supports 42+ languages.

471 downloads/mo

@thomasdavis/unsandbox(runCode)

Execute code with automatic language detection from shebang (e.g., #!/usr/bin/env python3). Simpler than executeCode when your code has a shebang line. Returns stdout, stderr, and detected language.

471 downloads/mo

@thomasdavis/unsandbox(executeCode)

Execute code synchronously in a secure sandbox. Supports 42+ languages including python, javascript, typescript, go, rust, c, cpp, java, ruby, and more. Returns stdout, stderr, and exit code.

471 downloads/mo

@thomasdavis/unsandbox(runCodeAsync)

Execute code asynchronously with automatic language detection from shebang. Returns a job_id immediately. Use getJob to check status and retrieve results.

471 downloads/mo

@thomasdavis/unsandbox(listJobs)

List all active jobs for your API key. Returns job IDs, statuses, and timestamps. Jobs are retained for 3 minutes after completion.

471 downloads/mo

@thomasdavis/unsandbox(getJob)

Get status and results of a specific async job. Returns status (pending, running, completed, cancelled, timeout), stdout, stderr, and timestamps.

471 downloads/mo

Test Scenarios

See how tool collections are tested with AI-generated scenarios. Real execution, real results.

Create Announcements Channel

Create a new channel for announcements in the Guild with ID `1234567890` and send a welcome message.

channel management

0% quality0 runs

Create Targeted Audience

Create a new audience for targeted marketing

0% quality0 runs

Update User Profile

Update the current user's profile with new settings

0% quality1 runs

model context protocol

works with every mcp client

Create a collection, get an MCP URL, add it to your client. Your AI gets instant access to all tools in the collection.

how it works

1

create a collection

Pick tools from the registry and group them into a collection. Public or private.

2

copy your mcp url

Each collection gets a unique URL that serves as an MCP endpoint.

3

add to your ai client

Paste the URL into your client's MCP config. Tools are available immediately.

setup by client

one command — that's it

claude mcp add my-tools \
  https://tpmjs.com/@username/collections/my-tools/mcp \
  -t http

tip

run /mcp in Claude Code to verify the server is connected. run claude mcp list to see all configured servers.

url format

https://tpmjs.com/@username/collections/slug/mcp

one url gives your ai access to 797+ tools

private collections

Private collections require a TPMJS API key. Add an Authorization: Bearer YOUR_KEY header to your config. Generate keys in Settings.

Publish Your AI Tool

Share your tool with the AI community. Automatic discovery, quality scoring, and integration with Vercel AI SDK, LangChain, and more.

Start with Our Package Generator

Create a production-ready TPMJS tool package in seconds with our CLI generator. Includes 2-3 tools, complete setup, and best practices.

npx @tpmjs/create-basic-tools

Quick Setup

Add one keyword to package.json and publish to NPM

Auto Discovery

Your tool appears on tpmjs.com within 15 minutes

Quality Metrics

Automatic scoring based on docs, downloads, and stars