mcpsync-cli

Unified MCP server configuration for all your coding agents.

Stop copying MCP server configs between Claude Code, Gemini CLI, OpenAI Codex, and other AI coding tools. Define your servers once, sync everywhere.

# Add a server once
mcp-sync add github --command npx --args "-y @modelcontextprotocol/server-github"

# Push to all your coding agents
mcp-sync push

# ✓ claude-code  (3 servers)
# ✓ codex        (3 servers)
# ✓ gemini-cli   (3 servers)
# ✓ roo-code     (3 servers)

Installation

npm install -g mcpsync-cli

Or use without installing:

npx mcpsync-cli init

Quick Start

# 1. Initialize config
mcp-sync init

# 2. Add your MCP servers
mcp-sync add github
# Interactive prompts guide you through setup

# 3. Sync to all detected agents
mcp-sync push

# 4. Check everything is working
mcp-sync doctor

Supported Agents

Agent	Status	Config Format	Notes
Claude Code	✅	JSON	Full support
OpenAI Codex	✅	TOML	Full support including tool filtering
Gemini CLI	✅	JSON	Full support including cwd, tool filtering, trust
Roo Code	✅	JSON	Full support with alwaysAllow, env var transform
Amp Code	✅	JSON	Full support with includeTools, httpUrl for HTTP
OpenCode	✅	JSON	Full support with disabledTools, autoApprove, env var transform
Kimi Code	🚧	JSON	Coming soon

Commands

mcp-sync init

Initialize a new configuration file.

mcp-sync init           # Create ~/.config/mcp-sync/config.yaml
mcp-sync init --force   # Overwrite existing config

mcp-sync add <n>

Add a new MCP server.

# Interactive mode
mcp-sync add github

# Non-interactive
mcp-sync add github \
  --type stdio \
  --command npx \
  --args "-y" "@modelcontextprotocol/server-github" \
  --env "GITHUB_TOKEN=\${GITHUB_TOKEN}"

# HTTP server
mcp-sync add context7 \
  --type http \
  --url "https://mcp.context7.com/mcp" \
  --env "CONTEXT7_API_KEY=\${CONTEXT7_API_KEY}"

mcp-sync remove <n>

Remove an MCP server.

mcp-sync remove github

mcp-sync list

List all configured servers.

mcp-sync list         # Table format
mcp-sync list --json  # JSON format

mcp-sync push [agent]

Sync configuration to agents.

mcp-sync push              # All installed agents
mcp-sync push claude-code  # Specific agent
mcp-sync push --dry-run    # Preview changes

mcp-sync agents

List detected agents and their status.

mcp-sync agents         # Table format
mcp-sync agents --json  # JSON format

mcp-sync doctor

Health check for your configuration.

mcp-sync doctor

Configuration

Configuration lives at ~/.config/mcp-sync/config.yaml:

version: "1"

defaults:
  timeout: 60
  autoApprove: false

servers:
  # Stdio server (local command)
  github:
    type: stdio
    command: npx
    args:
      - "-y"
      - "@modelcontextprotocol/server-github"
    env:
      GITHUB_PERSONAL_ACCESS_TOKEN: ${GITHUB_TOKEN}
    description: "GitHub API access"
    
    # Agent-specific overrides
    agents:
      codex:
        enabledTools:
          - list_repos
          - get_file_contents
        disabledTools:
          - delete_repo

  # HTTP server (remote)
  context7:
    type: http
    url: https://mcp.context7.com/mcp
    headers:
      CONTEXT7_API_KEY: ${CONTEXT7_API_KEY}
    description: "Documentation lookup"

# Optional: control which agents to sync to
agents:
  claude-code:
    enabled: true
    scope: user
  codex:
    enabled: true

# Optional: exclude servers from specific agents
exclusions:
  - server: github
    agent: roo-code
    reason: "Using different auth method"

Environment Variables

Use ${VAR_NAME} syntax to reference environment variables:

env:
  API_KEY: ${MY_API_KEY}           # Required
  API_KEY: ${MY_API_KEY:-default}  # With default value

Agent-Specific Settings

Some features are only supported by certain agents. Use the agents key to customize:

servers:
  myserver:
    type: stdio
    command: myserver
    agents:
      codex:
        # Codex supports tool filtering
        enabledTools: [tool1, tool2]
        disabledTools: [dangerous_tool]
      claude-code:
        # Different timeout for Claude
        timeout: 120

Programmatic Usage

import { ConfigManager, adapterRegistry } from 'mcpsync-cli';

// Load config
const manager = new ConfigManager();
const config = manager.get();

// Sync to specific adapter
const claudeAdapter = adapterRegistry.get('claude-code');
if (claudeAdapter) {
  await claudeAdapter.write(config);
}

// Detect installed agents
const installed = await adapterRegistry.getInstalled();
for (const adapter of installed) {
  console.log(`Found: ${adapter.displayName}`);
}

Development

# Install dependencies
npm install

# Run unit tests
npm run test:unit

# Run E2E tests in Docker
npm run test:e2e:docker

# Lint code
npm run lint

# Format code
npm run format

# Check formatting
npm run format:check

# Build
npm run build

Contributing

Contributions welcome! Areas we need help with:

• New adapters: Kimi Code
• Testing: Test coverage for adapters
• Documentation: Guides for specific workflows

Adding a New Adapter

1. Create src/adapters/your-agent.ts implementing BaseAdapter
2. Register in src/adapters/index.ts
3. Add tests in tests/adapters/your-agent.test.ts

See src/adapters/claude-code.ts for reference.

License

MIT