tooluniverse-cs-setup

Installs ToolUniverse and rebuilds its workflow library as a Claude Science skill.

1.7k|254|Updated Mar 3, 2025
One-click install
npx skills add https://github.com/mims-harvard/ToolUniverse --skill tooluniverse-cs-setup
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: tooluniverse-cs-setup
Source: https://github.com/mims-harvard/ToolUniverse/tree/main/skills/tooluniverse-cs-setup
Command: npx skills add https://github.com/mims-harvard/ToolUniverse --skill tooluniverse-cs-setup

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires tooluniverse.

What problem does it solve?

Setting up ToolUniverse inside Claude Science is non-trivial because the upstream project ships a Claude Code plugin with an MCP server and uvx, which Claude Science cannot load. This Skill performs the equivalent native installation: creating the conda environment, installing the tooluniverse pip package, and rebuilding the workflow library as a dynamically loaded skill.

Core Features & Use Cases

  • Environment and package setup: Creates a tooluniverse conda environment (Python 3.11) and installs or upgrades the tooluniverse pip package supplying 2500+ scientific tools.
  • Workflow bundle builder: Downloads the ToolUniverse GitHub tarball, parses every tooluniverse-* workflow SKILL.md, drops plugin/installer entries, and stages a publishable tooluniverse-research skill tree with SKILL.md, kernel.py, index.json, and workflows/*.md.
  • Publish and verify: Pushes the staged tree into the skill catalog via host.skills and validates the install by running a live tool call such as a PubChem compound lookup.
  • Use Case: After an upstream ToolUniverse release, rerun the four steps to upgrade the pip package and refresh all 133 bundled research workflows from the latest GitHub state.

Quick Start

Ask the agent to set up ToolUniverse for Claude Science by creating the conda environment, installing the tooluniverse package, and rebuilding the tooluniverse-research skill.

Frequently Asked Questions about tooluniverse-cs-setup

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I install ToolUniverse in Claude Science?

Create a conda environment named tooluniverse with Python 3.11, install the tooluniverse pip package into it, then run tu_build_research_bundle to stage the workflow library and publish it via host.skills. The skill walks through all four steps.

How do I update ToolUniverse workflows after a new release?

Rerun the pip install step to upgrade the tooluniverse package, then rerun the bundle build and publish steps. The builder fetches the latest GitHub tarball and reparses every tooluniverse-* workflow, replacing the old skill cleanly.

Does ToolUniverse on Claude Science need an MCP server or uv?

No. This setup replaces the upstream Claude Code plugin, MCP server, and uvx with a native pip installation. The tooluniverse package supplies the tools directly, and workflows are loaded through a dynamically published skill instead of slash commands.

Do ToolUniverse tools require API keys?

Most tools work without API keys. A few services such as NCBI, OncoKB, and NVIDIA unlock enhanced access when keys are provided; add them under Customize → Credentials and expose them in the tooluniverse conda environment.

Why does ToolUniverse caching fail in a sandboxed environment?

ToolUniverse defaults its cache to ~/.tooluniverse, which is read-only in the sandbox. The get_tu() helper redirects the cache to the workspace by setting TOOLUNIVERSE_CACHE_DIR, so no manual configuration is needed.