Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Research GitHub, GitLab, and Bitbucket repositories using DeepWiki MCP server. Use when exploring unfamiliar codebases, understanding project architecture, or asking questions about how a specific open-source project works. Provides AI-powered repo analysis and RAG-based Q&A about source code. NOT for fetching library API docs (use fetching-library-docs instead) or local files.
.claude/skills/aiskillstore-researching-with-deepwiki/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 2% | 0% |
| case-09 | ✗→✓ | ▲ Improved | -30% | 0% |
| case-14 | ✗→✓ | ▲ Improved | 44% | 0% |
| case-15 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-17 | ✗→✓ | ▲ Improved | -7% | 0% |
Research and understand codebases using the DeepWiki MCP server.
Add to your Claude Code settings:
bashclaude mcp add -s user -t http deepwiki https://mcp.deepwiki.com/mcp
Or add to settings.json:
json{ "mcpServers": { "deepwiki": { "type": "http", "url": "https://mcp.deepwiki.com/mcp" } } }
For private repos, use the Devin.ai endpoint with authentication:
json{ "mcpServers": { "deepwiki": { "type": "http", "url": "https://mcp.devin.ai/deepwiki/mcp", "headers": { "Authorization": "Bearer YOUR_DEVIN_API_KEY" } } } }
| Scenario | Use DeepWiki? | |----------|---------------| | Exploring unfamiliar open-source codebase | Yes | | Understanding project architecture | Yes | | Finding how a feature is implemented | Yes | | Generating documentation for a repo | Yes | | Working with your own local code | No - use Glob/Grep | | Quick file lookups in known structure | No - use Read tool |
Ask DeepWiki to analyze any public repository:
"Analyze the architecture of github.com/vercel/next.js"
"How is the routing system implemented in github.com/remix-run/react-router?"
"What design patterns are used in github.com/anthropics/anthropic-sdk-python?"Get insights into project organization:
"Explain the directory structure of github.com/langchain-ai/langchain"
"What are the main modules in github.com/fastapi/fastapi?"
"How are tests organized in github.com/pytest-dev/pytest?"Understand how specific features work:
"How does streaming work in github.com/openai/openai-python?"
"Where is authentication handled in github.com/better-auth/better-auth?"
"How are middleware implemented in github.com/honojs/hono?"DeepWiki can generate Mermaid diagrams:
"Generate an architecture diagram for github.com/prisma/prisma"
"Show the data flow in github.com/trpc/trpc"# Good - specific question
"How does the caching system work in github.com/vercel/swr?"
# Less effective - too broad
"Tell me about github.com/vercel/swr"# Good - targeted investigation
"Explain the src/core directory in github.com/pmndrs/zustand"
# Also good - feature-focused
"How is the middleware pattern implemented in github.com/pmndrs/zustand?""Compare how github.com/tanstack/query and github.com/vercel/swr handle cache invalidation""What patterns from github.com/shadcn-ui/ui should I follow for my component library?"1. "Explain the core concepts of github.com/honojs/hono"
2. "How do I add middleware in github.com/honojs/hono?"
3. "Show example route handlers from github.com/honojs/hono"1. "How does github.com/drizzle-team/drizzle-orm handle connection pooling?"
2. "What error types are thrown by github.com/drizzle-team/drizzle-orm?"1. "What's the contribution workflow for github.com/anthropics/claude-code?"
2. "How are issues labeled in github.com/anthropics/claude-code?"
3. "What testing patterns are used in github.com/anthropics/claude-code?"| Platform | URL Format | |----------|------------| | GitHub | github.com/owner/repo | | GitLab | gitlab.com/owner/repo | | Bitbucket | bitbucket.org/owner/repo |
Run: python3 scripts/verify.py
Expected: ✓ DeepWiki MCP server configured
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 55,096 | 50,255 | -9% | 1 | 1 | 0% | 4,935 | 7,888 | +60% | 0 | 0 | — |
case-02 | fail→fail | 41,978 | 44,250 | +5% | 1 | 1 | 0% | 5,878 | 8,698 | +48% | 0 | 0 | — |
case-03 | fail→fail | 24,930 | 25,790 | +3% | 1 | 1 | 0% | 3,507 | 6,018 | +72% | 0 | 0 | — |
case-04 | pass→pass | 14,200 | 5,117 | -64% | 1 | 1 | 0% | 1,675 | 2,104 | +26% | 0 | 0 | — |
case-05 | pass→pass | 6,567 | 8,156 | +24% | 1 | 1 | 0% | 1,077 | 1,718 | +60% | 0 | 0 | — |
case-06 | pass→pass | 8,157 | 9,784 | +20% | 1 | 1 | 0% | 1,555 | 1,974 | +27% | 0 | 0 | — |
case-07 | fail→pass | 14,276 | 2,362 | -83% | 1 | 1 | 0% | 1,645 | 1,678 | +2% | 0 | 0 | — |
case-08 | pass→pass | 13,776 | 6,917 | -50% | 1 | 1 | 0% | 1,157 | 1,626 | +41% | 0 | 0 | — |
case-09 | fail→pass | 16,649 | 7,937 | -52% | 1 | 1 | 0% | 2,545 | 1,784 | -30% | 0 | 0 | — |
case-10 | fail→fail | 21,138 | 20,909 | -1% | 1 | 1 | 0% | 2,581 | 3,866 | +50% | 0 | 0 | — |
case-11 | fail→fail | 22,603 | 8,890 | -61% | 1 | 1 | 0% | 2,263 | 2,397 | +6% | 0 | 0 | — |
case-12 | pass→pass | 19,582 | 18,122 | -7% | 1 | 1 | 0% | 3,101 | 3,484 | +12% | 0 | 0 | — |
case-13 | pass→pass | 20,995 | 12,051 | -43% | 1 | 1 | 0% | 2,506 | 2,422 | -3% | 0 | 0 | — |
case-14 | fail→pass | 7,524 | 8,090 | +8% | 1 | 1 | 0% | 1,171 | 1,691 | +44% | 0 | 0 | — |
case-15 | fail→pass | 19,154 | 8,383 | -56% | 1 | 1 | 0% | 1,915 | 1,850 | -3% | 0 | 0 | — |
case-16 | pass→pass | 18,952 | 7,100 | -63% | 1 | 1 | 0% | 2,443 | 1,507 | -38% | 0 | 0 | — |
case-17 | fail→pass | 13,153 | 2,828 | -78% | 1 | 1 | 0% | 1,872 | 1,741 | -7% | 0 | 0 | — |
case-18 | pass→pass | 16,103 | 8,411 | -48% | 1 | 1 | 0% | 1,604 | 1,811 | +13% | 0 | 0 | — |
case-19 | pass→pass | 15,540 | 2,024 | -87% | 1 | 1 | 0% | 1,584 | 1,533 | -3% | 0 | 0 | — |
case-20 | pass→pass | 7,343 | 7,122 | -3% | 1 | 1 | 0% | 1,284 | 1,604 | +25% | 0 | 0 | — |
case-21 | fail→pass | 11,319 | 9,105 | -20% | 1 | 1 | 0% | 1,183 | 2,015 | +70% | 0 | 0 | — |
case-22 | pass→pass | 21,953 | 9,799 | -55% | 1 | 1 | 0% | 2,922 | 2,003 | -31% | 0 | 0 | — |
case-23 | pass→pass | 11,455 | 7,482 | -35% | 1 | 1 | 0% | 2,014 | 2,444 | +21% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 23 cases were attempted. The headline lift of +26 percentage points is the difference between those two pass rates over the 23 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.