aio

AIO-MCP Server provides a modular, AI-powered MCP endpoint with integrations to services like GitLab, Jira, Confluence, YouTube, and more. It offers a suite of tools that can perform calendar management, Confluence queries, Jira operations, and various search and script utilities, all exposed through an MCP interface or via SSE (Server-Sent Events) HTTP endpoints. By enabling or disabling tool groups via ENABLE_TOOLS, you can tailor the server to only expose the capabilities you need. SSE mode allows real-time bi-directional communication over HTTP, making it suitable for web apps and remote clients. To start the server in SSE mode, configure the .env file or pass flags as shown in the example, then open the SSE endpoint and message endpoint to interact with the MCP server.

Prerequisites:

• Go 1.23.2 or higher
• API keys and tokens for the services you plan to use

1. Install via Smithery (Claude Desktop integration):

npx -y @smithery/cli install @athapong/aio-mcp --client claude

Smithery will guide you through interactive configuration and environment setup.

2. Manual Go installation:

# Set GOPATH and add bin to PATH (Linux/macOS example)
export GOPATH=$HOME/go
export PATH=$PATH:$GOPATH/bin

# Verify Go environment if needed
go env GOPATH

Install the server binary:

go install github.com/athapong/aio-mcp@latest

3. Prepare configuration file (example .env and mcp config):

Create a .env file with your credentials and feature toggles, e.g.:

ENABLE_TOOLS=
OPENAI_API_KEY=
GITLAB_HOST=
GITLAB_TOKEN=
ATLASSIAN_HOST=
ATLASSIAN_EMAIL=
ATLASSIAN_TOKEN=
GOOGLE_AI_API_KEY=
PROXY_URL=
OPENAI_BASE_URL=
QDRANT_HOST=
QDRANT_API_KEY=
QDRANT_PORT=
USE_OLLAMA_DEEPSEEK=
DEEPSEEK_API_KEY=
OLLAMA_URL=http://localhost:11434

Then, use a configuration snippet like this (Claude Desktop config example):

{
  "mcpServers": {
    "aio-mcp": {
      "command": "aio-mcp",
      "args": ["-env", "/path/to/.env", "-sse", "-sse-addr", ":8080", "-sse-base-path", "/mcp"]
    }
  }
}

To override environment values directly in the config:

{
  "mcpServers": {
    "aio-mcp": {
      "command": "aio-mcp",
      "env": {
        "ENABLE_TOOLS": "",
        "OPENAI_API_KEY": "",
        "GITLAB_TOKEN": "",
        "GITLAB_HOST": "",
        "QDRANT_HOST": "",
        "QDRANT_API_KEY": "",
        "PROXY_URL": "",
        "OPENAI_API_KEY": "",
        "GOOGLE_TOKEN_FILE": "",
        "GOOGLE_CREDENTIALS_FILE": "",
        "ATLASSIAN_TOKEN": "",
        "BRAVE_API_KEY": "",
        "QDRANT_PORT": "",
        "ATLASSIAN_HOST": "",
        "ATLASSIAN_EMAIL": "",
        "USE_OPENROUTER": "",
        "DEEPSEEK_API_KEY": "",
        "OPENROUTER_API_KEY": "",
        "DEEPSEEK_API_BASE": "",
        "USE_OLLAMA_DEEPSEEK": "",
        "OLLAMA_URL": ""
      }
    }
  }
}

4. Run the server using the generated configuration.

Tips and common issues:

• Ensure your environment variables are correctly set in .env or via the config; missing keys can disable tool groups or cause auth failures.
• If SSE mode is used, ensure the chosen port is open and not blocked by a firewall.
• When using DeepSeek or Ollama, verify that the local services are running and reachable at the default endpoints (e.g., http://localhost:11434).
• The ENABLE_TOOLS setting accepts a comma-separated list of tool groups to enable; leave empty to enable all by default.
• For Google, Jira, Confluence integrations, ensure proper host URLs and tokens are configured in the environment.
• If you encounter connectivity issues, check PROXY_URL and network egress rules.
• The command shown uses -env to load credentials from a file; you can also inject values directly via the env map in the config.
• SSE_BASE_PATH defaults to /mcp; adjust as needed to fit your web app routing.