MCP Lite Wrappers

Lightweight MCP wrapper plugins that consolidate multiple tools into single-action tools, reducing Claude Code context usage by ~80%.

Problem

MCP servers expose many individual tools, each consuming context tokens. For example, supabase-mcp exposes 20+ tools, using ~11k tokens just for tool definitions.

Solution

Wrap multiple tools into a single tool with an action parameter. One tool, one description, massive token savings.

Before: 20 tools × ~500 tokens = ~10,000 tokens
After:  1 tool  × ~1,000 tokens = ~1,000 tokens
Savings: ~80%

Packages

Installation

supabase-lite-mcp

npm install supabase-lite-mcp

Add to your Claude Code MCP config (~/.claude/settings.json or project .mcp.json):

{
  "mcpServers": {
    "supabase": {
      "command": "npx",
      "args": ["supabase-lite-mcp"],
      "env": {
        "SUPABASE_ACCESS_TOKEN": "your-access-token",
        "SUPABASE_PROJECT_REF": "your-project-ref"
      }
    }
  }
}

linear-lite-mcp

npm install linear-lite-mcp

Add to your Claude Code MCP config:

{
  "mcpServers": {
    "linear": {
      "command": "npx",
      "args": ["linear-lite-mcp"],
      "env": {
        "LINEAR_API_KEY": "lin_api_xxxxx"
      }
    }
  }
}

Get your Linear API key from Linear Settings > API.

chrome-lite-mcp

npm install chrome-lite-mcp

Add to your Claude Code MCP config:

{
  "mcpServers": {
    "chrome": {
      "command": "npx",
      "args": ["chrome-lite-mcp"],
      "env": {
        "CHROME_PATH": "/path/to/chrome",
        "CHROME_HEADLESS": "true"
      }
    }
  }
}

Environment variables (optional):

• CHROME_PATH - Path to Chrome executable (auto-detected if not set)
• CHROME_HEADLESS - Set to "false" for visible browser (default: headless)

context7-lite-mcp

npm install context7-lite-mcp

Add to your Claude Code MCP config:

{
  "mcpServers": {
    "context7": {
      "command": "npx",
      "args": ["context7-lite-mcp"],
      "env": {
        "CONTEXT7_API_KEY": "your-api-key"
      }
    }
  }
}

Environment variables (optional):

• CONTEXT7_API_KEY - Context7 API key from context7.com/dashboard (optional - lower rate limits without it)

Usage

Supabase

All actions go through a single supabase tool:

// List projects
{ action: "list_projects" }

// Execute SQL
{ action: "execute_sql", payload: { query: "SELECT * FROM users" } }

// Deploy edge function
{ action: "deploy_edge_function", payload: { slug: "my-func", code: "..." } }

Available Actions

Projects & Organizations

• list_projects, get_project, create_project, pause_project, restore_project
• list_organizations, get_organization

Cost Management

• get_cost, confirm_cost

Database

• execute_sql, list_tables, list_extensions
• list_migrations, apply_migration

Monitoring

• get_logs, get_advisors

Project Info

• get_project_url, get_publishable_keys, generate_typescript_types

Edge Functions

• list_edge_functions, get_edge_function, deploy_edge_function

Branching

• create_branch, list_branches, delete_branch
• merge_branch, reset_branch, rebase_branch

Storage

• list_storage_buckets, get_storage_config, update_storage_config

Documentation

• search_docs

Linear

All actions go through a single linear tool:

// Create an issue
{ action: "create_issue", payload: { title: "Fix bug", teamId: "TEAM-ID" } }

// Search issues
{ action: "search_issues", payload: { query: "login bug", limit: 10 } }

// Get user's teams
{ action: "get_user_teams" }

Linear Available Actions

Issues

• create_issue, update_issue, get_issue, search_issues
• get_user_issues, get_team_issues, get_project_issues

Comments

• add_comment, get_comments

Teams & Projects

• get_teams, get_team, get_projects, get_project

Labels

• get_labels, create_label, update_label

Users

• get_viewer, get_users, get_user_teams, get_user_projects

Issue Relations

• link_issues, get_issue_relations

Attachments

• add_attachment, get_attachments

Workflow & Milestones

• get_workflow_states, get_milestones, create_milestone, update_milestone

Chrome

All actions go through a single chrome tool:

// Navigate to a URL with auto-capture (saves .md, .html, .png)
{ action: "navigate_page", payload: { url: "https://example.com", autoCapture: true, outputDir: "./captures" } }

// Click an element
{ action: "click", payload: { selector: "#submit-btn" } }

// Fill a form field
{ action: "fill", payload: { selector: "#email", value: "test@example.com" } }

// Select option from dropdown
{ action: "select", payload: { selector: "#country", value: "US" } }

// Extract content in different formats
{ action: "extract", payload: { selector: ".article", format: "markdown" } }

// Take a screenshot
{ action: "take_screenshot", payload: { fullPage: true } }

// Show browser (switch from headless to visible)
{ action: "show_browser" }

// Set browser profile
{ action: "set_profile", payload: { name: "work", userDataDir: "~/.chrome-profiles/work" } }

Chrome Available Actions

Input Automation

• click, drag, fill, fill_form, handle_dialog, hover, press_key, upload_file, select

Navigation

• navigate_page (with auto-capture support), new_page, list_pages, select_page, close_page, wait_for

Emulation

• emulate, resize_page

Performance

• performance_start_trace, performance_stop_trace, performance_analyze_insight

Network

• get_network_request, list_network_requests

Extraction & Debugging

• extract (text/html/markdown), get_attr, evaluate_script, take_screenshot, take_snapshot, list_console_messages, get_console_message

Browser Control

• show_browser, hide_browser, browser_mode

Profile Management

• set_profile, get_profile

Context7

All actions go through a single context7 tool:

// Resolve library ID
{ action: "resolve_library_id", payload: { libraryName: "next.js" } }

// Get library docs
{ action: "get_library_docs", payload: {
  context7CompatibleLibraryID: "/vercel/next.js",
  topic: "routing",
  tokens: 5000
} }

Context7 Available Actions

Library Resolution

• resolve_library_id - Search for libraries and get Context7-compatible IDs

Documentation

• get_library_docs - Fetch up-to-date, version-specific documentation with optional topic filtering

Architecture

Follows the superpowers-chrome pattern:

1. Single tool with action enum parameter
2. Switch/dispatch based on action
3. Proxy calls to underlying API (Supabase Management API, Linear GraphQL API)

packages/
├── supabase-lite/
│   ├── src/
│   │   ├── index.ts      # MCP server entry
│   │   ├── actions.ts    # Action dispatcher
│   │   ├── types.ts      # Zod schemas
│   │   └── client/       # API client
│   └── package.json
├── linear-lite/
│   ├── src/
│   │   ├── index.ts      # MCP server entry
│   │   ├── actions.ts    # Action dispatcher
│   │   ├── types.ts      # Zod schemas
│   │   └── client/       # GraphQL client
│   └── package.json
└── chrome-lite/
    ├── src/
    │   ├── index.ts      # MCP server entry
    │   ├── actions.ts    # Action dispatcher
    │   ├── types.ts      # Zod schemas
    │   └── browser/      # Puppeteer wrapper
    └── package.json

Development

# Install dependencies
npm install

# Build
npm run build

# Run tests
npm test

# Type check
npm run typecheck

Tech Stack

• TypeScript
• Zod for validation
• MCP SDK (@modelcontextprotocol/sdk)
• puppeteer-core (for chrome-lite)

License

MIT

References

• superpowers-chrome - Pattern reference
• supabase-mcp - Supabase MCP being wrapped
• Linear GraphQL API - Linear API documentation
• Chrome DevTools MCP - Chrome DevTools reference
• Puppeteer - Browser automation library
• MCP SDK - Model Context Protocol SDK