jilebi MCP server
Commit Oct 20, 2026Extends AI assistants with custom tools, resources, and prompts through a modular, plugin-based architecture.
Also known as Datron/jilebi
- Stars
- 4
- Tools
- —
- Downloads, 30 days
- —
- Latest version
- —
- Last commit
- Oct 20, 2026
- License
- AGPL-3.0
What is jilebi MCP?
jilebi extends AI assistants with custom tools, resources, and prompts through a modular, plugin-based architecture.
Summary generated from the project's README
About jilebi MCP server
Extending AI assistants with custom tools, resources, and prompts is the primary function of jilebi. This MCP server uses a modular, plugin-based architecture to allow language models to interact with external systems. It is designed for developers working in environments like Claude Desktop, Cursor, VS Code, Zed, and Codex CLI. The server runs locally via stdio and can be installed on Windows, Linux, and macOS using platform-specific scripts.
Users can expand functionality by adding plugins such as memory or sequential-thinking. The system includes a granular permission model to manage network access and file system operations, ensuring security when plugins request access to specific URLs or directories. Developers can also build their own plugins using a TOML manifest to define custom tools, read-only resources, and conversation templates. The server supports persistent state management, allowing plugins to save and retrieve data across different sessions.
What you can do with it
- 1Expand AI capabilities with pluginsA developer can add specialized functionality like memory or sequential thinking to their AI assistant.
- 2Secure file and network accessDevelopers can restrict what parts of the file system or which web domains a plugin is allowed to touch.
- 3Manage persistent plugin dataYou can instruct an AI to save information that remains available even after the session ends.
Written from the project's README. Every use case is checked against a quote from it.
README
View on GitHubA powerful, plugin-based MCP (Model Context Protocol) server that extends AI assistants with custom tools, resources, and prompts.
Table of Contents
- Overview
- Download
- Installation
- Host Configuration
- Plugin System
- Creating Plugins
- Logging and Debugging
- Links
Overview
Jilebi is an MCP server implementation that enables AI assistants to interact with external systems through a modular plugin architecture. It provides a standardized way to expose tools, resources, and prompts to language models while maintaining security through a granular permission system.
Key Features:
- Cross-platform support (Windows, Linux, macOS)
- Plugin-based extensibility
- Secure permission model for file system and network access
- Persistent state management for plugins
- Compatible with major AI development environments
Quick Install
The fastest way to install Jilebi is using our installation scripts:
Linux & macOS
Using curl:
curl -fsSL https://mcp.jilebi.ai/install.sh | bash
Using wget:
wget -qO- https://mcp.jilebi.ai/install.sh | bash
Windows (PowerShell)
irm https://mcp.jilebi.ai/install.ps1 | iex
Or with the full command:
Invoke-RestMethod https://mcp.jilebi.ai/install.ps1 | Invoke-Expression
What the installer does
- Detects your platform and architecture automatically
- Downloads the appropriate binary
- Installs to
~/.jilebi/bin(Unix) or%LOCALAPPDATA%\jilebi\bin(Windows) - Adds Jilebi to your PATH
- Optionally installs recommended plugins (memory, sequential-thinking)
Uninstall
To completely remove Jilebi from your system:
Linux & macOS
Using curl:
curl -fsSL https://mcp.jilebi.ai/uninstall.sh | bash
Using wget:
wget -qO- https://mcp.jilebi.ai/uninstall.sh | bash
Windows (PowerShell)
irm https://mcp.jilebi.ai/uninstall.ps1 | iex
Or with options:
# Skip confirmation prompts
irm https://mcp.jilebi.ai/uninstall.ps1 -OutFile uninstall.ps1; .\uninstall.ps1 -Force
# Keep plugins and data
irm https://mcp.jilebi.ai/uninstall.ps1 -OutFile uninstall.ps1; .\uninstall.ps1 -KeepData
What the uninstaller removes
- Jilebi binary and installation directory
- PATH entry from shell profile (Unix) or user environment (Windows)
- Optionally: plugins, data, and configuration files
Manual Download
| Platform | Architecture | Download |
|---|---|---|
| Windows 10/11 | x86_64 | Download |
| Linux | x86_64 | Download |
| macOS | Apple Silicon | Download |
| macOS | Intel | Download |
Manual Installation
Windows
# 1. Unzip the downloaded file to your desired location
# 2. Install recommended plugins
.\jilebi.exe plugins add memory
.\jilebi.exe plugins add sequential-thinking
# 3. Start the MCP server
.\jilebi.exe stdio
Linux
# 1. Unzip the downloaded file to your desired location
# 2. Grant execution access
chmod +x jilebi
# 3. Install recommended plugins
./jilebi plugins add memory
./jilebi plugins add sequential-thinking
# 4. Start the MCP server
./jilebi stdio
macOS
# 1. Unzip the downloaded file to your desired location
# 2. Grant execution access
chmod +x jilebi
# 3. Remove quarantine attribute (required for unverified developers)
xattr -d com.apple.quarantine /path/to/jilebi
# 4. Install xz dependency
brew install xz
# 5. Install recommended plugins
./jilebi plugins add memory
./jilebi plugins add sequential-thinking
# 6. Start the MCP server
./jilebi stdio
Find more plugins at github.com/datron/jilebi-plugins
Host Configuration
Jilebi integrates with various AI development environments. Below are configuration examples for supported hosts.
Claude Desktop
Config Location:
- Windows:
%APPDATA%\Claude\claude_desktop_config.json - macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
{
"mcpServers": {
"jilebi": {
"command": "jilebi",
"args": ["stdio"],
"env": {}
}
}
}
Claude Code
Add to .claude.json:
{
"mcpServers": {
"jilebi": {
"command": "jilebi",
"args": ["stdio"],
"env": {}
}
}
}
Cursor
Edit ~/.cursor/mcp.json:
{
"mcpServers": {
"jilebi": {
"command": "jilebi",
"args": ["stdio"],
"env": {}
}
}
}
VS Code
Add to your mcp.json:
{
"servers": {
"jilebi": {
"type": "stdio",
"command": "jilebi",
"args": ["stdio"]
}
}
}
Zed
Edit ~/.config/zed/settings.json:
{
"context_servers": {
"jilebi": {
"command": "jilebi",
"args": ["stdio"],
"enabled": true,
"source": "custom",
"env": {}
}
}
}
Codex CLI
Add to ~/.codex/config.toml:
[mcp_servers.jilebi]
command = "jilebi"
args = ["stdio"]
Opencode
Edit ~/.config/opencode.json:
{
"$schema": "https://opencode.ai/config.json",
"mcp": {
"jilebi": {
"type": "local",
"command": ["jilebi", "stdio"],
"enabled": true,
"environment": {}
}
}
}
Plugin System
Plugins extend Jilebi with custom functionality through three main components:
| Component | Description |
|---|---|
| Tools | Functions that models can invoke to interact with external systems |
| Resources | Read-only data sources for context (files, schemas, documentation) |
| Prompts | Pre-defined conversation templates with argument interpolation |
Managing Plugins
# Install a plugin
jilebi plugins add <plugin-name>
# Create a new plugin
jilebi plugins create
# Setup/reconfigure a plugin
jilebi plugins setup <plugin-name>
# View plugin logs
jilebi plugins log <plugin-id>
Creating Plugins
Quick Start
jilebi plugins create
Follow the prompts to bootstrap a new plugin project.
Plugin Manifest
Plugins are defined using a TOML manifest file. Here is an example structure:
# Metadata
name = "my-plugin"
version = "1.0.0"
homepage = "github.com/user/my-plugin"
creator = "Your Name"
contact = "you@example.com"
# Environment Variables
[env]
API_URL = { schema = { type = "string" }, default = "https://api.example.com" }
# Secrets (no defaults allowed)
[secrets]
API_KEY = { schema = { type = "string" } }
# Tools
[tools.my-tool]
name = "my-tool"
description = "Description of what this tool does"
input_schema = { type = "object", properties = { param = { type = "string" } }, required = ["param"] }
function = "my_tool_function"
[tools.my-tool.permissions]
hosts = ["https://api.example.com"]
# Resources
[resources.my-resource]
name = "my-resource"
description = "Description of this resource"
mime_type = "application/json"
function = "get_resource"
# Prompts
[prompts.my-prompt]
name = "my-prompt"
description = "A helpful prompt template"
arguments = [
{ name = "topic", description = "The topic to discuss", required = true }
]
messages = [
{ role = "user", content = { type = "text", content = "Tell me about {{topic}}" } }
]
Permission Types
Plugins can request the following permissions:
| Permission | Description |
|---|---|
hosts |
Allowed host domains for network requests |
urls |
Specific allowed URLs |
read_files |
File paths the plugin can read |
write_files |
File paths the plugin can write |
read_dirs |
Directories the plugin can read from |
write_dirs |
Directories the plugin can write to |
Use "user_defined" to prompt the user for a custom value during setup.
Built-in Functions
Plugins have access to:
State Management:
setState(env, key, value)- Persist data across sessionsgetState(env, key)- Retrieve persisted datadeleteState(env, key)- Remove persisted data
File System (Deno API):
- Directory operations:
mkdir,readDir - File operations:
readFile,writeFile,readTextFile,writeTextFile - File manipulation:
stat,copyFile,rename,remove - Temporary files:
makeTempDir,makeTempFile - And more...
Logging and Debugging
Log Locations
| Platform | Path |
|---|---|
| Windows | C:\Users\<user>\AppData\Roaming\jilebi\jilebi-server\data\logs |
| Linux | ~/.local/share/jilebi-server/logs/ |
| macOS | ~/Library/Application Support/ai.jilebi.jilebi-server/logs |
Viewing Logs
# View Jilebi server logs
jilebi log
# View plugin-specific logs
jilebi plugins log <plugin-id>
Writing Logs in Plugins
console.log("Info message");
console.warn("Warning message");
console.error("Error message");
console.debug("Debug message");
Links
jilebi MCP questions
Which AI development environments are supported?
Jilebi integrates with Claude Desktop, Claude Code, Cursor, VS Code, Zed, Codex CLI, and Opencode.