A Python SDK for building AI research systems that need to orchestrate scientific tools across proteins, drugs, genomics, and literature databases. You get three calling patterns (portable dict API, function API, or direct instantiation), built-in caching for deterministic calls, and batch execution for parallel workflows. The tool discovery is embedding-based, which is helpful when you're not sure which of the 100+ tools you actually need. Main gotcha is remembering to call load_tools() before anything else, and tool names are case-sensitive. Best for building reproducible research pipelines or benchmarking systems where you need programmatic access to UniProt, ChEMBL, PubMed, AlphaFold predictions, and similar scientific APIs without writing individual integrations yourself.

skills.sh installs
399
GitHub stars
1.8K
last updated
Aug 21, 2026

Install

$ npx -y skills add mims-harvard/tooluniverse --skill tooluniverse-sdk --agent claude-code

Installs into .claude/skills of the current project.

Files

ToolUniverse Python SDK

3 calling patterns -- start with pattern 1:

  1. tu.run({"name": ..., "arguments": ...}) -- single tool call, dict API (most portable)
  2. tu.tools.ToolName(param=value) -- function API (recommended for interactive use)
  3. Direct class instantiation -- advanced, bypasses caching/hooks

Installation

pip install tooluniverse              # Standard
pip install tooluniverse[embedding]   # Embedding search (GPU)
pip install tooluniverse[all]         # All features
export OPENAI_API_KEY="sk-..."  # Required for LLM tool search
export NCBI_API_KEY="..."       # Optional

Quick Start

from tooluniverse import ToolUniverse

tu = ToolUniverse()
tu.load_tools()  # REQUIRED before any tool call

# Find tools
tools = tu.run({"name": "Tool_Finder_Keyword", "arguments": {"description": "protein structure", "limit": 10}})

# Execute (dict API)
result = tu.run({"name": "UniProt_get_entry_by_accession", "arguments": {"accession": "P05067"}})

# Execute (function API)
result = tu.tools.UniProt_get_entry_by_accession(accession="P05067")

Core Patterns

Batch Execution

Sponsors

Details

Repository
github.com/mims-harvard/tooluniverse
Categories
Backend & APIs, DevOps & CI/CD, AI & Agent Building, Data Science & ML, Automation & Workflows, Python
First seen
Jun 3, 2026

More from mims-harvard/tooluniverse

All 68
1aliyun avataralibabacloud-das-agentaliyun/alibabacloud-aiops-skillsDiagnose and manage Alibaba Cloud databases through natural language. Use when users need to troubleshoot database performance issues (high CPU, slow queries, a395skills.sh installs
2davila7 avatarstripe-integrationdavila7/claude-code-templatesGet paid from day one. Payments, subscriptions, billing portal, webhooks, metered billing, Stripe Connect. The complete guide to implementing Stripe correctly, 392skills.sh installs
3mongodb avataratlas-stream-processingmongodb/agent-skillsManages MongoDB Atlas Stream Processing (ASP) workflows. Handles workspace provisioning, data source/sink connections, processor lifecycle operations, debugging392skills.sh installs
4sarvamai avatarspeech-to-textsarvamai/skillsWrite correct Sarvam Saaras STT code for 23 Indic languages — REST modes, Batch API with diarization, and WebSocket streaming gotchas. Use this skill when build389skills.sh installs
5senpi-ai avatarsenpi-entrypointsenpi-ai/senpi-skillsOnboard AI agents into the Senpi trading platform on Hyperliquid. Walks through account creation, API key setup, MCP server configuration, skill discovery, and 389skills.sh installs
6auth0 avatarauth0-java-mvc-commonauth0/agent-skillsUse when adding Auth0 login, logout, or callback handling to a Java Servlet web application. Integrates com.auth0:mvc-auth-commons for server-side Java apps — u388skills.sh installs