agentcfg 0.2.0

CLI tool for deploying and managing AI coding agent configurations (MCP servers, skills, instructions) across multiple providers.

agent-config

A Python CLI tool for deploying and managing AI coding agent configurations — MCP servers, instruction files (AGENTS.md, CLAUDE.md), skills, and runtime memory — across multiple providers and machines.

Keep your configs in version control. Run agent-config deploy to push them to every tool.

Install

pip install agentcfg

Or with uv:

uv tool install agentcfg

Quick Start

# Scaffold a new config directory
agent-config init

# Edit the generated files to match your setup
# (mcp_providers.json, agents.json, instruction_providers.json …)

# Deploy everything
agent-config deploy

Commands

Command	Description
agent-config init	Scaffold a new config repo with starter templates
agent-config deploy	Deploy manifest entries (files/dirs) to their destinations
agent-config pull	Pull memory files back from deployed locations into the repo
agent-config sync	Deploy then run agents sync across all LLM providers
agent-config mcp deploy	Write MCP server configs to all registered provider files
agent-config mcp sync	Additive MCP sync — add only servers missing from disk
agent-config mcp show	Show MCP configuration status per provider
agent-config instructions deploy	Copy instruction files (AGENTS.md etc.) to each provider
agent-config instructions show	Show instruction file status per provider
agent-config skills deploy	Deploy all registered skills to their target agents
agent-config skills sync	Deploy only skills not yet installed
agent-config skills show	Show skill deployment status
agent-config providers list	List registered MCP providers
agent-config providers add	Add a new MCP provider to the registry
agent-config providers remove	Remove an MCP provider
agent-config providers set-path	Update a provider's config file path

Global options

Flag	Description
--config-dir DIR / -C DIR	Root of your config repo (default: cwd or $AGENT_CONFIG_DIR)
--version / -V	Print version and exit

Filtering and dry-run

Most commands accept --providers p1 p2 … to limit to specific providers, and --dry-run to preview changes without writing anything.

Config Files

File	Purpose
manifest.json	Maps source files/dirs to their deploy destinations
agents.json	MCP server definitions and integration toggles
mcp_providers.json	Registry of MCP provider config paths and formats
instruction_providers.json	Registry of instruction file destinations per provider
skills_registry.json	Registry of skills (local directories or skills-cli packages)

Prerequisites

Tool	Purpose
Python 3.8+	Runtime for agent-config
uv	Package manager
just	Task runner (optional, for justfile recipes)
Node.js + npx	Required only for installer: skills-cli manifest entries
Git	Required for pull command (git add) and submodules

Environment Variables

Variable	Description
AGENT_CONFIG_DIR	Override the default config directory (equivalent to --config-dir)

Supported MCP Providers

Out of the box: VS Code, GitHub Copilot CLI, Claude Desktop, OpenCode, Cursor, Windsurf, JetBrains Junie.

Add custom providers with agent-config providers add.

Cross-Platform

Works on Windows (PowerShell), macOS, and Linux. Path expansion supports both %USERPROFILE% (Windows) and ~ / $HOME (POSIX).