jilebi

Jilebi is a plugin-based MCP server that extends AI assistants with external tools, resources, and prompts. It exposes a modular architecture where plugins can add Tools (functions models can call to interact with external systems), Resources (read-only data sources for context), and Prompts (pre-defined conversation templates with argument interpolation). You can start the server in stdio mode, which exposes the local process as an MCP endpoint for client integrations. Once running, you can manage plugins (install, create, configure, and view logs), and then compose prompts and tools to enable your AI assistant to perform complex tasks with external capabilities. The included host configuration snippets show how to wire Jilebi into popular environments like Claude, Cursor, VS Code, and others, using the standard command and stdio workflow.

To use Jilebi effectively, install or load plugins that provide the tools you need (e.g., memory, sequential-thinking, or domain-specific plugins). You can then expose Tools through the MCP to your language model interfaces, enabling actions such as file operations, network requests, or data transformations via well-defined prompts and environment-controlled capabilities. Start the server with the stdio entry point and ensure your client or IDE is configured to communicate with the jilebi MCP endpoint using the jilebi stdio command and arguments as shown in the host configuration examples.

Prerequisites:

• A supported operating system (Linux, macOS, or Windows)
• Basic command line tools (curl or wget on UNIX-like systems)
• Network access to download installer binaries if using the quick install method

Linux & macOS (fast install):

1. Open a terminal.
2. Run the installation script via curl: curl -fsSL https://jilebi.ai/install.sh | bash (This detects your platform/architecture, downloads the appropriate binary, and installs it to your PATH.)
3. Alternatively, install via wget: wget -qO- https://jilebi.ai/install.sh | bash
4. After installation, verify that jilebi is on your PATH and accessible: jilebi --version
5. Install recommended plugins (optional but suggested): jilebi plugins add memory jilebi plugins add sequential-thinking
6. Start the MCP server in stdio mode: jilebi stdio

Windows (PowerShell):

1. Open PowerShell.
2. Install via the installer script: irm https://jilebi.ai/install.ps1 | iex or the full command: Invoke-RestMethod https://jilebi.ai/install.ps1 | Invoke-Expression
3. Confirm installation by checking the binary is in PATH and run: jilebi --version
4. Install recommended plugins (optional): jilebi plugins add memory jilebi plugins add sequential-thinking
5. Start the MCP server: jilebi stdio

Windows/Linux/macOS uninstall (if needed):

• Uninstall scripts are provided by the installer and can be run to remove binaries and configuration. See the official Jilebi documentation for exact commands.

Tips and common considerations:

• The core server runs in stdio mode and is designed to be wired into IDEs and toolchains via the provided host configuration examples (Claude, Cursor, VS Code, Zed, Codex CLI, Opencode, etc.).
• Plugins extend functionality; use jilebi plugins add <plugin-name> to install and jilebi plugins log <plugin-id> to view activity.
• If you enable memory or sequential-thinking plugins, ensure you understand memory usage and potential security implications for sensitive data.
• Environment configuration can be adjusted per host; for most setups the default env is empty but you can populate it with API keys, endpoints, or other secrets as needed via the plugin TOML manifest or host-specific mcp.json entries.
• When integrating with editors or CLIs, use the standard command and args (stdio) as shown in the host configuration snippets to ensure compatibility.
• For troubleshooting, consult the logs via jilebi plugins log or start the server with verbose logging if supported by your binary.