Shipkit get mcps
Skill stefan-stepzero/shipkit/install/skills/shipkit-get-mcps
Shipkit — AI-assisted product development framework for Claude Code. Skills, agents, and workflows for shipping MVPs fast.
npx -y skills add stefan-stepzero/shipkit --skill shipkit-get-mcpsAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
One thing to look at
- 1 stars1 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
What its author says it does
Copied from the file, not written here
Discover and install MCP servers for enhanced Claude capabilities
SKILL.md
8.0 KB, as published. Nobody here has run it
shipkit-get-mcps - MCP Discovery & Installation
Purpose: Discover and install MCP (Model Context Protocol) servers to extend Claude's capabilities.
Core value: User knows to ask — Claude handles discovery and installation.
Why This Skill Exists
You don't naturally know what MCPs exist or that they could help.
| Without Skill | With Skill |
|---|---|
| User doesn't know MCPs exist | User invokes /get-mcps |
| User manually searches, edits config | Claude searches, presents options, installs |
| User misses useful capabilities | Claude suggests relevant MCPs |
When to Invoke
User says:
- "Get MCPs for database"
- "Is there an MCP for Playwright?"
- "I need browser automation"
- "Find MCPs", "Search MCPs"
- "Add an MCP for Supabase"
- "What MCPs are available?"
- "Show me popular MCPs"
- "Recommend an MCP"
Auto-suggest when:
- User struggles with a task that an MCP could help with
- User asks about external service integration
For catalog questions: Load references/popular-mcps.md and present options.
Prerequisites
Required:
- Node.js installed (most MCPs use npx)
- Internet connection for discovery
Not required:
- No manifest or registry account needed
- Claude searches on behalf of user
Completion Tracking
When installing an MCP, create tasks:
TaskCreate: "Search/discover MCP options"TaskCreate: "Read existing .mcp.json (merge, don't overwrite)"TaskCreate: "Detect platform (Windows/Mac/Linux) for correct config"TaskCreate: "Write updated .mcp.json"TaskCreate: "Post-install: restart reminder + prerequisite warnings"
TaskUpdate each task to in_progress when starting it, completed when done.
Writing .mcp.json must MERGE with existing MCPs, not overwrite. Platform detection determines whether to use npx directly or cmd wrapper.
Process
Step 1: Understand Need
When user mentions a need:
User: I need to interact with my database
Claude: I can search for MCPs that provide database integration.
What database are you using? (Postgres, Supabase, MongoDB, etc.)
If user invokes directly with query:
User: /get-mcps supabase
→ Skip to Step 2
Step 2: Find MCPs
First, check the curated catalog:
→ Read references/popular-mcps.md for common MCPs (databases, browsers, dev tools, etc.)
If user asks "what's available" or wants to browse: → Present options from the catalog grouped by category
If not in catalog, USE PARALLEL SUBAGENTS to search multiple registries:
Launch these Agent subagents IN PARALLEL (single message, multiple tool calls):
1. OFFICIAL REGISTRY AGENT (subagent_type: "Explore")
Prompt: "Search for '[query]' MCP server on official registry.
WebFetch: registry.modelcontextprotocol.io
Return: matching MCPs with name, description, install command."
2. COMMUNITY DIRECTORY AGENT (subagent_type: "Explore")
Prompt: "Search for '[query]' MCP server on community sites.
WebSearch: '[query] MCP server site:mcp.so'
WebSearch: '[query] site:mcpservers.org'
Return: matching MCPs with name, description, source URL."
3. GITHUB SEARCH AGENT (subagent_type: "Explore")
Prompt: "Search for '[query]' MCP server on GitHub.
WebSearch: '[query] MCP server site:github.com'
Return: matching repos with name, stars, description, install command."
Why parallel: Network fetches to different registries are independent - run simultaneously for faster discovery.
After agents return:
- Merge results from all sources
- Deduplicate (same MCP may appear in multiple registries)
- Rank by relevance and popularity (stars, official status)
- Present consolidated list to user
Step 3: Present Options
Format results clearly:
Found 2 MCPs for "supabase":
[1] @supabase/mcp-server (official)
Database queries, auth, storage, realtime
Source: npm
[2] supabase-mcp (community)
Lightweight Supabase integration
Source: github.com/user/supabase-mcp
Install one? Enter number, or [s] to search again
Step 4: Install Selected MCP
Read current config:
cat .mcp.json 2>/dev/null || echo "{}"
Add new MCP entry:
For Unix/Mac:
{
"mcpServers": {
"supabase": {
"command": "npx",
"args": ["-y", "@supabase/mcp-server"]
}
}
}
For Windows (detect via platform):
{
"mcpServers": {
"supabase": {
"command": "cmd",
"args": ["/c", "npx", "-y", "@supabase/mcp-server"]
}
}
}
Write updated config:
- Merge with existing mcpServers
- Preserve other MCPs
- Validate JSON before writing
Step 5: Post-Install
✓ Added supabase to .mcp.json
⚠ Restart Claude Code to activate the new MCP.
The MCP will be available after restart. You'll have access to:
- Database query tools
- Auth management
- Storage operations
Step 6: Handle Edge Cases
MCP requires prerequisites
⚠ playwright MCP requires browser binaries.
Run this first:
npx playwright install
Then restart Claude Code.
No .mcp.json exists
No .mcp.json found. Creating one...
✓ Created .mcp.json with supabase MCP
Create minimal config:
{
"mcpServers": {
"supabase": { ... }
}
}
MCP already installed
supabase is already in .mcp.json
Current config:
command: npx
args: ["-y", "@supabase/mcp-server"]
[u] Update args [r] Remove [b] Back
No results found
No MCPs found for "quantum computing".
Try:
- Broader terms: "database", "api", "automation"
- Check mcp.so directly: https://mcp.so
- The capability might not have an MCP yet
Commands
| Command | Description |
|---|---|
/get-mcps | Interactive discovery prompt |
/get-mcps <query> | Search by keyword |
/get-mcps list | Show installed MCPs from .mcp.json |
/get-mcps remove <name> | Remove MCP from config |
Discovery Sources
| Source | URL | Type |
|---|---|---|
| MCP Registry | registry.modelcontextprotocol.io | Official |
| mcp.so | mcp.so | Community |
| mcpservers.org | mcpservers.org | Curated |
| GitHub | github.com (search) | Source repos |
Platform Detection
Detect platform and adjust config:
Windows:
command: "cmd"
args: ["/c", "npx", "-y", "<package>"]
Unix/Mac:
command: "npx"
args: ["-y", "<package>"]
Check platform via environment or ask user if unclear.
Context Files This Skill Reads
.mcp.json— Current MCP configuration
Context Files This Skill Writes
.mcp.json— Add/update MCP entries
Write Strategy: MERGE
- Read existing config
- Add new MCP to mcpServers
- Preserve existing MCPs
- Write back
When This Skill Integrates with Others
Relationship: Standalone Utility
This skill operates independently. It helps users extend Claude's capabilities.
Complements:
/shipkit-project-context— Stack detection might suggest MCPs
Sibling:
/shipkit-get-skills— Same pattern for skill discovery
<!-- SECTION:after-completion -->
After Completion
Guardrails Check: Before moving to next task, verify:
- Persistence - Has important context been saved to
.shipkit/? - Prerequisites - Does the next action need a spec or plan first?
- Session length - Long session? Consider
/shipkit-work-memoryfor continuity.
Natural capabilities (no skill needed): Implementation, debugging, testing, refactoring, code documentation.
Suggest skill when: User needs to make decisions, create persistence, or check project status.
<!-- /SECTION:after-completion -->Success Criteria
- User can discover MCPs through conversation
- Search covers official and community sources
- Installation handles platform differences
- Restart reminder always shown
- Prerequisites flagged when known