@conduitai-app/mcp-server

@conduitai-app/mcp-server

MCP server for conduitAI — search, browse, and install AI tools directly from your AI assistant.

conduitAI is the trusted marketplace for MCP servers, agents, skills, prompts, and plugins. This MCP server lets Claude, Cursor, Windsurf, and other AI clients interact with the full conduitAI catalog natively via the Model Context Protocol.

Quick Start

Claude Code

Add to your MCP config (~/.claude/.mcp.json):

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

Cursor / Windsurf

Add the same command and args to your editor's MCP settings panel.

VS Code (Copilot)

Add to your .vscode/mcp.json:

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

Docker

docker run -it --rm ghcr.io/conduitai-app/mcp-server

Authentication

On first use, call the authenticate tool. It opens your browser to sign in to conduitAI — no manual token management needed.

If you've already authenticated with the conduitAI CLI (conduitai auth login), the MCP server picks up your existing session automatically from ~/.conduit/config.json.

Environment Variable Authentication

For CI/CD or headless environments, set the CONDUITAI_TOKEN environment variable:

{
  "mcpServers": {
    "conduitai": {
      "command": "npx",
      "args": ["-y", "@conduitai-app/mcp-server"],
      "env": {
        "CONDUITAI_TOKEN": "your-token-here"
      }
    }
  }
}

Tools

authenticate

Sign in to conduitAI via browser. Opens an OAuth flow — no manual token management needed.

auth_status

Check your current authentication status. Returns whether you're signed in and your account info.

search_catalog

Search the conduitAI catalog across all asset types.

Parameter	Type	Required	Description
query	string	Yes	Search query
type	string	No	Filter by type: mcp_server, mcp_client, agent, skill, prompt, plugin
sort	string	No	Sort order: quality, stars, downloads, recent
verified_only	boolean	No	Only show verified assets
limit	number	No	Max results, 1-20 (default: 10)

get_asset_details

Get full details for any asset by its slug, including description, quality grade, GitHub stats, tags, and install methods.

Parameter	Type	Required	Description
slug	string	Yes	Asset slug (e.g. "filesystem-mcp-server")

browse_catalog

Browse assets by type, sorted by popularity or recency.

Parameter	Type	Required	Description
type	string	Yes	Asset type: mcp_server, mcp_client, agent, skill, prompt, plugin
sort	string	No	Sort order: popular, recent
limit	number	No	Max results, 1-20 (default: 10)

install_asset

Install an asset to your local environment. For MCP servers, this updates your Claude Code / Cursor configuration. For skills, this writes skill files to ~/.claude/skills/.

Parameter	Type	Required	Description
slug	string	Yes	Asset slug to install
scope	string	No	"global" (default) or "project"
dry_run	boolean	No	Preview changes without applying

uninstall_asset

Remove a previously installed asset, cleaning up config files and local state.

Parameter	Type	Required	Description
slug	string	Yes	Asset slug to uninstall

list_installed

List assets currently installed on this machine.

Parameter	Type	Required	Description
type	string	No	Filter by asset type

Examples

Once installed, you can ask your AI assistant:

• "Search conduitAI for PostgreSQL MCP servers"
• "Show me the most popular agents on conduitAI"
• "Install the filesystem MCP server from conduitAI"
• "What conduitAI assets do I have installed?"
• "Find verified MCP servers for database access, sorted by quality"
• "Uninstall the slack-mcp-server"

How It Works

The MCP server is a thin protocol adapter that translates MCP tool calls into HTTPS requests to conduitAI's API infrastructure. It supports two transport modes:

• stdio (default) — for local AI clients like Claude Code and Cursor
• Streamable HTTP — for remote deployments (e.g. the hosted instance at mcp.conduitai.app)

Your requests go through the same rate-limited, quota-enforced endpoints as the conduitAI web app and CLI. All install specs are cryptographically signed to prevent tampering.

Configuration

Environment Variable	Description	Default
CONDUITAI_TOKEN	Authentication token (for CI/headless use)	—
CONDUITAI_API	API base URL override (for development)	conduitAI production API

Token and session data are stored in ~/.conduit/config.json with 0600 permissions (owner-only read/write).

Security

• Signed install specs — all install operations use cryptographically signed specs fetched from the server to prevent tampering
• Restricted file permissions — config files are written with 0600 (owner read/write only)
• SSRF protection — the API client validates redirect targets against a trusted domain allowlist and enforces HTTPS
• Input validation — all tool inputs are validated with Zod schemas before processing
• No secrets in stdout — MCP protocol uses stdout for JSON-RPC; all debug logging goes to stderr
• Rate limiting — the HTTP transport enforces per-user rate limits with standard X-RateLimit-* headers

Requirements

• Node.js 18 or later
• A conduitAI account (free at conduitai.app)

Development

git clone https://github.com/conduitai-app/mcp-server.git
cd mcp-server
npm install
npm run dev      # Run with tsx (hot reload)
npm run build    # Compile TypeScript
npm test         # Run tests

TypeScript

This package is written in TypeScript and ships compiled JavaScript in dist/. Type definitions are included. The source uses ES modules ("type": "module").

Related

• @conduitai-app/cli — conduitAI command-line interface
• conduitai.app — conduitAI web marketplace
• Model Context Protocol — MCP specification

Contributing

Contributions are welcome. Please open an issue first to discuss what you'd like to change.

• Fork the repository
• Create your feature branch (git checkout -b feat/my-feature)
• Make your changes and add tests
• Run npm run build && npm test to verify
• Open a pull request

License

MIT