Three Steps for Learning and Visualization with Claude Code
A 3-skill toolkit for learning new technologies without leaving your workflow. Escalate from quick terminal diagrams, through interactive browser exploration, to full AI-generated learning packages.
| Step | Skill | Environment | When to use |
|---|
| 1 | /learn-toolkit:visualize | Terminal | Quick flowchart or architecture diagram, stay in CLI |
| 2 | /learn-toolkit:playground | Browser (HTML) | Compare alternatives, explore parameters interactively |
| 3 | /learn-toolkit:learn | Terminal + Browser + Web | All 3 steps combined: ASCII diagram + interactive playground + NotebookLM package with study guide |
Install
Option A: Plugin install (recommended)
/plugin marketplace add yodem/learn-toolkit
/plugin install learn-toolkit@learn-toolkit-marketplace
This installs all 3 skills, configures MCP servers (via env vars), and handles updates automatically.
Then add your API keys to ~/.zshrc (or ~/.bashrc):
export TAVILY_API_KEY="your-key-here" # https://tavily.com (free)
export EXA_API_KEY="your-key-here" # https://exa.ai
Restart Claude Code. Done.
/learn-toolkit:visualize and /learn-toolkit:playground work immediately — no API keys needed. Keys are only required for /learn-toolkit:learn's search backends.
Option B: Paste repo URL
Paste this URL into Claude Code and it reads CLAUDE.md to walk you through setup:
https://github.com/yodem/learn-toolkit
Also works with Cursor, Windsurf, and other AI tools that read CLAUDE.md.
Option C: Manual setup
Click to expand manual steps
See Prerequisites below for required API keys and tools.
Step 1: Install the skills
# All 3 skills
mkdir -p ~/.claude/skills/learn/references ~/.claude/skills/visualize ~/.claude/skills/playground
cp plugins/learn-toolkit/skills/learn/SKILL.md ~/.claude/skills/learn/SKILL.md
cp plugins/learn-toolkit/skills/learn/references/*.md ~/.claude/skills/learn/references/
cp plugins/learn-toolkit/skills/visualize/SKILL.md ~/.claude/skills/visualize/SKILL.md
cp plugins/learn-toolkit/skills/playground/SKILL.md ~/.claude/skills/playground/SKILL.md
Step 2: Configure MCP servers
Add to ~/.claude/settings.json under "mcpServers" — note the ${VAR} references, not literal keys:
{
"mcpServers": {
"tavily": {
"type": "url",
"url": "https://mcp.tavily.com/mcp/?tavilyApiKey=${TAVILY_API_KEY}"
},
"exa": {
"type": "url",
"url": "https://mcp.exa.ai/mcp?exaApiKey=${EXA_API_KEY}&tools=web_search_exa,web_search_advanced_exa,get_code_context_exa,crawling_exa,company_research_exa,people_search_exa,deep_researcher_start,deep_researcher_check"
}
}
}
Then add your actual keys to your shell profile (~/.zshrc or ~/.bashrc):
export TAVILY_API_KEY="tvly-your-key-here"
export EXA_API_KEY="your-exa-key-here"
This way your config file contains no secrets and can be safely shared or committed.
Step 3: Install NotebookLM MCP (optional)
# See https://github.com/nicholasgriffintn/notebooklm-mcp
nlm login
Step 4: Restart Claude Code
/exit
claude
Tavily Agent Skills (recommended)
Install Tavily's agent skills to give /learn-toolkit:learn a CLI-based fallback and add standalone search/extract/crawl/research capabilities:
npx skills add tavily-ai/skills --yes
curl -fsSL https://cli.tavily.com/install.sh | bash
tvly login # opens browser for OAuth, or: tvly login --api-key tvly-YOUR_KEY
This installs 7 skills as slash commands:
| Skill | Slash Command | CLI Command | Purpose |
|---|
tavily-search | /tavily-search | tvly search "query" --json | Web search with LLM-optimized results |
tavily-extract | /tavily-extract | tvly extract "url" | Extract content from specific URLs |
tavily-crawl | /tavily-crawl | tvly crawl "url" --output-dir ./docs/ | Crawl site sections to local markdown |
tavily-map | /tavily-map | tvly map "url" | Discover URLs on a domain |
tavily-research | /tavily-research | tvly research "topic" --json | Deep multi-source research with citations |
tavily-cli | /tavily-cli | tvly --help | Unified CLI reference |
tavily-best-practices | /tavily-best-practices | — | Integration patterns and guidance |
Workflow escalation: search → extract → map → crawl → research
Integration with /learn-toolkit:learn: The learn workflow auto-detects Tavily CLI alongside the MCP server. If MCP is available, it uses MCP for search and CLI skills for extract/crawl. If only the CLI is installed, skills handle all Tavily operations. Either path provides full Tavily coverage.
The Three Skills
Step 1: /learn-toolkit:visualize — ASCII Diagrams in Terminal
1270011
upstash
severity1