JSPM

@tankpkg/mcp-server

0.15.7
  • ESM via JSPM
  • ES Module Entrypoint
  • Export Map
  • Keywords
  • License
  • Repository URL
  • TypeScript Types
  • README
  • Created
  • Published
  • Downloads 1056
  • Score
    100M100P100Q115269F
  • License MIT

MCP server for Tank - scan and publish AI agent skills from your editor

Package Exports

    This package does not declare an exports field, so the exports above have been automatically detected and optimized by JSPM instead. If any package subpath is missing, it is recommended to post an issue to the original package (@tankpkg/mcp-server) to support the "exports" field. If that is not possible, create a JSPM override to customize the exports field for this package.

    Readme

    @tankpkg/mcp-server

    MCP (Model Context Protocol) server for Tank — manage AI agent skills directly from your editor. Full CLI parity: every tank command is available as an MCP tool.

    Features

    Authentication

    • login — Authenticate with Tank via GitHub OAuth
    • logout — Clear stored credentials
    • whoami — Show current user info

    Project Setup

    • init-skill — Create skills.json and SKILL.md scaffold

    Publishing & Discovery

    • publish-skill — Publish a skill to the Tank registry (with dry-run support)
    • search-skills — Search the Tank registry for skills
    • skill-info — Get detailed information about a specific skill

    Installation & Management

    • install-skill — Install a skill with SHA-512 verification
    • update-skill — Update skills within semver range
    • remove-skill — Remove a skill and clean up lockfile

    Security & Verification

    • scan-skill — Scan any directory for security issues (skills.json not required)
    • verify-skills — Verify lockfile integrity
    • audit-skill — Show security scan results and verdict
    • skill-permissions — Display per-skill permission summary

    Agent Integration

    • link-skill — Symlink a skill into an agent workspace
    • unlink-skill — Remove a skill symlink

    Diagnostics

    • doctor — Check config, auth, registry connectivity, and Node.js version

    Installation

    Claude Code

    Add to .claude/settings.json or your project's .mcp.json:

    {
      "mcpServers": {
        "tank": {
          "command": "npx",
          "args": ["-y", "@tankpkg/mcp-server"]
        }
      }
    }

    Cursor

    Add to ~/.cursor/mcp.json:

    {
      "mcpServers": {
        "tank": {
          "command": "npx",
          "args": ["-y", "@tankpkg/mcp-server"]
        }
      }
    }

    VS Code (Copilot)

    Add to .vscode/mcp.json:

    {
      "servers": {
        "tank": {
          "command": "npx",
          "args": ["-y", "@tankpkg/mcp-server"]
        }
      }
    }

    Authentication

    The MCP server shares authentication with the Tank CLI. If you've already run tank login, you're authenticated!

    Alternatively, set the TANK_TOKEN environment variable:

    {
      "mcpServers": {
        "tank": {
          "command": "npx",
          "args": ["-y", "@tankpkg/mcp-server"],
          "env": {
            "TANK_TOKEN": "tank_your_token_here"
          }
        }
      }
    }

    Usage Examples

    Once configured, talk to your AI agent naturally:

    "Initialize a new skill in this directory"
    "Scan my skill for security issues"
    "Publish my skill to Tank"
    "Install @acme/code-review"
    "Update all my skills"
    "Search Tank for testing skills"
    "Show permissions for my installed skills"
    "Run tank doctor to check my setup"
    "Audit @acme/code-review for security issues"
    "Link this skill to my Claude workspace"

    Development

    # Install dependencies
    bun install
    
    # Build
    bun build
    
    # Run tests
    bun test
    
    # Start the server (stdio mode)
    bun start

    License

    MIT