Skip to main content

Mac Agents Manager

Web UI and CLI for managing macOS LaunchAgents, with AI Chat powered by Ollama.

CI PyPI PyPI Downloads Python License

Mac Agents Manager lets you create, view, edit, start, stop, and reload user LaunchAgents from a web dashboard or the mam CLI — no manual plist editing or raw launchctl commands required. The dashboard includes an AI Chat mode powered by a local Ollama model for natural-language agent management.

What's New (2.1.0)

  • Restart Running KeepAlive — One button restarts every KeepAlive agent that is currently running, skipping stopped and scheduled agents
  • Honest bulk-action reporting — Partial failures in the new bulk restart surface as an error with counts, not a blanket success

Previously (2.0.1)

  • AI Chat Assistant — Natural-language control of LaunchAgents with Apply/Cancel confirmation before mutations
  • Session persistence — Chat sessions saved to ~/.mac_agents_manager/chat/ and restorable from the session picker
  • Ollama auto-lifecycle — Auto-start, crash recovery, and model auto-pull (same pattern as Chronometry)
  • Safety hardening — Server-side confirmation resolves only unresolved matching pending actions; fallback rewrites unstructured mutation claims

Features

  • Web Dashboard — Modern browser UI to browse, create, edit, and control LaunchAgents on localhost
  • AI Chat Assistant — Natural-language control with explicit Apply/Cancel confirmation before mutations
  • Unified CLI — Single mam command for all operations (list, create, start, stop, logs, delete)
  • LaunchAgent Lifecycle — Load, unload, start, stop, restart, and delete agents with one command
  • Bulk Actions — Start all stopped KeepAlive agents, or restart all running ones, from the sidebar
  • Agent Creation — Create new keepalive or scheduled LaunchAgents from the CLI or web UI
  • Log Viewer — Tail stdout/stderr logs for any agent with follow mode
  • Auto-Start — Install Mac Agents Manager itself as a LaunchAgent so the dashboard starts at login
  • Localhost Only — Binds to 127.0.0.1 for single-user, local-only access

How It Works

┌────────────────────────────────────────────────────────────────┐
│                           Your Mac                             │
│                                                                │
│   🖥️ Web Dashboard (localhost:8081)    ⌨️ CLI (mam)            │
│   ├─ Browse all LaunchAgents           ├─ mam list             │
│   ├─ Create / Edit agents              ├─ mam create           │
│   ├─ Start / Stop / Restart            ├─ mam start <label>    │
│   ├─ View logs                         ├─ mam logs <label>     │
│   └─ AI Chat (Ollama)                  └─ mam open             │
│           │                                    │               │
│           └────────────┬───────────────────────┘               │
│                        ▼                                       │
│      ┌───────────────────────────────────────────────┐         │
│      │      ~/Library/LaunchAgents/*.plist            │         │
│      │  User LaunchAgent property list files          │         │
│      └───────────────────────┬───────────────────────┘         │
│                              │                                 │
│                              ▼                                 │
│      ┌───────────────────────────────────────────────┐         │
│      │            launchctl (macOS)                   │         │
│      │  Load · Unload · Start · Stop · Status         │         │
│      └───────────────────────────────────────────────┘         │
│                                                                │
│      Everything runs locally. Nothing leaves your machine.     │
└────────────────────────────────────────────────────────────────┘

Quick Start

Prerequisites

  • macOS (uses macOS-specific launchctl APIs)
  • Python 3.10+ — check with python3 --version. If below 3.10, install it:
    brew install python@3.10
    
  • Ollama — local LLM runtime (required for AI Chat)
# Install Ollama
brew install ollama

# Start Ollama as a background service (auto-starts at login)
brew services start ollama

# Pull the text model used by AI Chat
ollama pull qwen3.5:4b

Note: If you skip Ollama setup, everything except AI Chat works normally. The chat health indicator will show "disconnected" until Ollama is available. MAM also attempts to auto-start Ollama and auto-pull the model on first chat use, but pre-installing is recommended for the best experience.

Install

# From PyPI
pip3 install mac-agents-manager-ai

# Or with uv
uv pip install mac-agents-manager-ai

# Or in a dedicated virtual environment
mkdir -p ~/.mac_agents_manager
python3 -m venv ~/.mac_agents_manager/venv
source ~/.mac_agents_manager/venv/bin/activate
pip install mac-agents-manager-ai

Run

# Install as macOS service (auto-start at login)
mam service install

# Or start manually
mam serve

# Open the dashboard
mam open

The dashboard is at http://localhost:8081.

Verify

# Check service status
mam service status

# List all LaunchAgents
mam list

# Check version
mam --version

AI Chat (Web Dashboard)

The dashboard right panel has two tabs: IDE (the classic editor) and AI Chat.

  • No service selected — Global suggested prompts: "Summarize all services", "Service statistics", "List failed agents", etc.
  • Service selected — Context-aware prompts: "Change schedule", "Rename", "Start", "Stop", "View logs", and all other IDE-equivalent actions.
  • Confirmation-first mutations — The assistant proposes changes and requires Apply/Cancel before execution.
  • Safer confirmations — Server-side confirmation only executes unresolved matching pending actions.
  • Session restore — Chat sessions persist to ~/.mac_agents_manager/chat/ and can be resumed from the session picker.
  • Fallback safety — If the model claims a mutation without structured action payload, the response is rewritten to a safe retry instruction.

Ollama Configuration

Variable Default Description
MAM_OLLAMA_BASE_URL http://localhost:11434 Ollama API base URL
MAM_OLLAMA_MODEL qwen3.5:4b Model used for AI Chat

Chat API endpoints:

Endpoint Method Description
/api/chat/health GET Ollama + model readiness check
/api/chat/send POST Send a message and get AI response
/api/chat/confirm POST Apply or cancel a pending mutation
/api/chat/history GET Retrieve chat history for a session
/api/chat/sessions GET List available chat sessions
/api/chat/clear POST Clear chat history for a session

CLI Reference

mam                              # Start web server (default)
mam serve [-p PORT] [--debug]    # Start web server
mam service install              # Install as LaunchAgent (auto-start at login)
mam service uninstall            # Uninstall the LaunchAgent
mam service start|stop|restart   # Control the LaunchAgent service
mam service status               # Show service status
mam list                         # List all agents with status
mam show <label>                 # Show agent details and plist XML
mam create <name> -c CAT -s SCRIPT [-t keepalive|scheduled]
mam load <label>                 # Register agent with launchd
mam unload <label>               # Unregister agent from launchd
mam start <label>                # Start a loaded agent
mam stop <label>                 # Stop a running agent
mam restart <label>              # Restart an agent
mam delete <label> [-y]          # Unload and delete an agent
mam start-all                    # Start all stopped keepalive agents
mam logs <label> [-f] [--stderr] [-n N]  # View agent logs
mam open                         # Open dashboard in browser
mam --version                    # Version info

Architecture

src/mac_agents_manager/
├── __init__.py       # Package version
├── cli.py            # Unified CLI (argparse)
├── app.py            # Flask web dashboard and routes (IDE + Chat API)
├── models.py         # LaunchAgent parsing, serialization, and UI data
├── launchctl.py      # Thin wrapper around launchctl commands
├── ollama_chat.py    # Ollama lifecycle, system prompt, action parsing, chat engine
├── chat_history.py   # Chat session persistence (~/.mac_agents_manager/chat/)
├── templates/        # HTML templates (Jinja2)
└── static/           # CSS styles

Runtime Directory

Chat sessions and AI state are stored in ~/.mac_agents_manager/:

~/.mac_agents_manager/
├── chat/                    # Chat session history (JSON files)
│   ├── session_abc123.json  # Individual session with messages + pending actions
│   └── ...
└── venv/                    # Virtual environment (if installed per Quick Start)

Environment Variables

Variable Default Description
MAM_PORT 8081 Port to listen on
FLASK_DEBUG off Set to 1 or true to enable Flask debug mode
MAM_LABEL_PREFIXES Comma-separated extra label prefixes to include (e.g. com.myorg.,com.acme.)
MAM_OLLAMA_BASE_URL http://localhost:11434 Ollama API base URL
MAM_OLLAMA_MODEL qwen3.5:4b Model used for AI Chat

Security

This tool binds to 127.0.0.1 only and is designed for single-user, localhost use. AI Chat mutations require explicit user confirmation. See SECURITY.md for the full security model and vulnerability reporting instructions.

Contributing

Contributions are welcome! Please read CONTRIBUTING.md before submitting a pull request. This project follows the Contributor Covenant Code of Conduct.

Development

# Clone and install in dev mode
git clone https://github.com/pkasinathan/mac-agents-manager-ai.git
cd mac-agents-manager-ai
make dev

# Run linter
make lint

# Auto-format
make format

# Run tests
make test

# All quality checks
make check

Release Process

Use the release checklist in RELEASE.md before publishing.

Uninstall

mam service uninstall
pip3 uninstall mac-agents-manager-ai

To also remove all chat history and local data:

rm -rf ~/.mac_agents_manager

Warning: Deleting ~/.mac_agents_manager permanently removes all chat sessions and local configuration. This cannot be undone.

License

Apache License 2.0 — see LICENSE.

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

mac_agents_manager_ai-2.1.0.tar.gz (85.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

mac_agents_manager_ai-2.1.0-py3-none-any.whl (72.5 kB view details)

Uploaded Python 3

File details

Details for the file mac_agents_manager_ai-2.1.0.tar.gz.

File metadata

  • Download URL: mac_agents_manager_ai-2.1.0.tar.gz
  • Upload date:
  • Size: 85.4 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.14.6

File hashes

Hashes for mac_agents_manager_ai-2.1.0.tar.gz
Algorithm Hash digest
SHA256 b12ad59db85e52bf8d70f342fc870f2ca600d2b112f52b15206e83262eb4910d
MD5 ebfdc315d2ca18cab1e8fda73eef6014
BLAKE2b-256 7bcc4a1a38d59adc508607f9de3afdc70e6b3148284d0b820012ae7ef6f7ce13

See more details on using hashes here.

File details

Details for the file mac_agents_manager_ai-2.1.0-py3-none-any.whl.

File metadata

File hashes

Hashes for mac_agents_manager_ai-2.1.0-py3-none-any.whl
Algorithm Hash digest
SHA256 b5bcf8bee6bc0dd6028ef84eb359845d2e9017b21acfd7b38dcfe3a0ff186d92
MD5 e79acb6bf0bf9b3b16550d43f09956ee
BLAKE2b-256 643fc08e5c816e640ff44f16d18f38c693e1d0355c774da41cc2fb1be2b7c448

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page