← Discover MCPs and Agents
o
AgentAI & MLGitHub

opencow

One task, one agent, delivered. The open-source platform for task-driven autonomous AI agents.OpenCow assigns an autonomous AI agent to every task — features, campaigns, reports, audits — and delivers them in parallel. Full context. Full control. Every department. 🐄

Links

README

From the repo.

English · 简体中文

OpenCow

OpenCow

One Task. One Agent. Delivered.

The open-source platform for task-driven autonomous AI.
Every task becomes an autonomous agent — campaigns, reports, features, audits
ship in parallel. For every team.

License: Apache-2.0 Electron React TypeScript PRs Welcome

How It Works · Features · Quick Start · Architecture · Contributing

Website · X (Twitter) · Download


OpenCow Issues Preview

How It Works

Task In. Result Out.

StepWhat happens
CreateWrite a task, not a prompt. Describe the deliverable — a campaign, a report, a feature, an audit. OpenCow links full context: project files, prior work, related tasks.
DispatchOne task, one agent. Each task gets a dedicated agent with full context — project knowledge, team playbooks, organizational standards. 15 tasks, 15 agents, in parallel.
DeliverAgents research, draft, build, and publish — autonomously. Real-time progress, instant notifications, approval gates at every step. Review. Ship.

The Platform

Everything to turn your task list into a parallel AI workforce.

Task-to-Agent Pipeline

Built-in task tracker where every task becomes an agent. Break projects into sub-tasks, each with a dedicated agent. Your task list IS your delivery plan.

1 Task = 1 Agent · Sub-task hierarchy · Auto-linked context

Agent Intelligence

Equip agents with your org's knowledge, standards, and tools. Skills, playbooks, integrations — every agent follows your processes.

Custom skills · 6 capability types · Auto-sync standards

Agent Command Center

Real-time dashboard. Track every agent's progress and actions. Approve deliverables. One screen, full visibility.

Live monitoring · Task-linked status · Approval gates

Work from Anywhere

Dispatch agents from Telegram, Discord, WeChat, or Lark. Schedule recurring workflows. Get notified via webhooks.

4 IM platforms · Natural language · 7 schedule types


Everything. Built In.

No plugins. No integrations to configure. Every capability your AI workforce needs.

Task & Agent Core
Task Tracker · Agent Dashboard
Live Monitor · Multi-Project

Intelligence
Intelligence Hub · Marketplace
Built-in Browser · Artifacts

Automation
Scheduling · Webhooks
Notifications · Live Preview

Command & Control
IM Command · Terminal
Command Palette · Themes


Built Different

The design decisions that define OpenCow.

PrincipleWhat it means
1:1Task → AgentOne task, one agent. Full context. Full traceability. Zero ambiguity.
Local & PrivateEverything runs on your machine. Zero telemetry. Zero cloud. Your data never leaves.
15+Parallel AgentsDeliver 15+ tasks simultaneously. Suspend and resume any agent without losing context.
4-LayerDeep Context EngineEvery agent inherits organizational knowledge, project context, team standards, and task-specific instructions.

Quick Start

From source (for contributors)

Prerequisites: Node.js >= 18 and pnpm >= 9

# Clone the repository
git clone https://github.com/OpenCowAI/opencow.git
cd opencow

# Install dependencies
pnpm install

# Launch in development mode (HMR enabled)
pnpm dev

Download the app

Grab the latest release from opencow.ai/download — free, open source, ready in 60 seconds.


Tech Stack

LayerChoiceWhy
DesktopElectron 40Cross-platform, native-grade experience
UIReact 19 + Tailwind CSS 4Concurrent rendering, utility-first styling
LanguageTypeScript (strict, zero any)End-to-end type safety
StateZustandLightweight reactive stores with row-level subscriptions
Buildelectron-vite (Vite)Sub-second HMR, triple-target build
DatabaseSQLite via KyselyLocal-first, type-safe schema with 35+ migrations
Terminalxterm.js + WebGLHardware-accelerated terminal rendering
EditorMonaco EditorVS Code-grade editing experience
TestingVitest + React Testing LibraryFast, modern test runner
AIClaude Agent SDK + Codex SDK + MCPMulti-engine — choose Claude or Codex as your AI engine, with Model Context Protocol

Architecture

OpenCow follows a hardened Electron architecture with strict process isolation:

┌─────────────────────────────────────────────────────┐
│                   Renderer Process                   │
│  ┌───────────┐  ┌──────────┐  ┌──────────────────┐  │
│  │  React 19  │  │  Zustand  │  │  271 Components  │  │
│  │ Components │  │  Stores   │  │  68 Hooks        │  │
│  └─────┬─────┘  └────┬─────┘  └────────┬─────────┘  │
│        └──────────────┴─────────────────┘            │
│                        │ IPC (contextBridge)         │
├────────────────────────┼────────────────────────────┤
│                    Main Process                      │
│  ┌──────────┐  ┌──────────┐  ┌───────────────────┐  │
│  │  DataBus  │  │ Services │  │  Native Modules   │  │
│  │  Events   │  │  (47+)   │  │  SQLite · PTY     │  │
│  └──────────┘  └──────────┘  └───────────────────┘  │
│  ┌──────────┐  ┌──────────┐  ┌───────────────────┐  │
│  │  Claude   │  │  Bot     │  │  Capability       │  │
│  │  Agent SDK│  │  Gateway  │  │  Center           │  │
│  └──────────┘  └──────────┘  └───────────────────┘  │
└─────────────────────────────────────────────────────┘

Security model:

  • contextIsolation: true — renderer cannot access Node.js
  • nodeIntegration: false — no direct module loading
  • Type-safe IPC bridge via contextBridge.exposeInMainWorld
  • Sandboxed file access with explicit path allowlists
  • Separate data directories for dev (~/.opencow-dev) and production (~/.opencow)

Project Structure

opencow/
├── electron/                  # Main process
│   ├── main.ts               # App entry point
│   ├── preload.ts            # Secure IPC bridge
│   ├── services/             # 47+ backend service modules
│   │   ├── capabilityCenter/ # AI capability management
│   │   ├── schedule/         # Cron automation engine
│   │   ├── messaging/        # Multi-channel messaging
│   │   ├── git/              # Git integration layer
│   │   └── ...
│   ├── database/             # SQLite schema & migrations
│   ├── sources/              # Hook, task, and stats data sources
│   └── ipc/                  # IPC channel handlers
├── src/
│   ├── renderer/             # React application
│   │   ├── components/       # 32 feature modules, 271 components
│   │   ├── hooks/            # 68 custom React hooks
│   │   ├── stores/           # 18 Zustand domain stores
│   │   ├── lib/              # Utilities & helpers
│   │   └── locales/          # i18n (en-US, zh-CN)
│   └── shared/               # Cross-process types & utilities
├── tests/                    # Vitest test suite
├── docs/                     # 520+ design docs & proposals
└── resources/                # App icons & tray assets

Development

Scripts

CommandWhat it does
pnpm devStart with HMR
pnpm buildCompile all targets
pnpm previewBuild + launch in production mode
pnpm packageBuild + package .app (fast, for testing)
pnpm package:dmgBuild + package .dmg installer
pnpm typecheckTypeScript strict check
pnpm lintESLint
pnpm formatPrettier
pnpm testRun all tests
pnpm test:watchWatch mode

Build Targets

electron-vite compiles three independent TypeScript targets:

TargetEntryOutput
Mainelectron/main.tsout/main/
Preloadelectron/preload.tsout/preload/
Renderersrc/renderer/index.htmlout/renderer/

Path Aliases

  • @src/renderer (renderer only)
  • @sharedsrc/shared (all targets)

Code Standards

  • TypeScript strict mode, zero any usage
  • Prettier — no semicolons, single quotes, 100 char width, 2-space indent
  • ESLint — strict rules with React Hooks plugin
  • Conventional Commitsfeat:, fix:, perf:, refactor:, etc.

Packaging

# macOS .app (for local testing, fast)
pnpm package
# Output: dist/mac-arm64/OpenCow.app

# macOS .dmg (for distribution)
pnpm package:dmg
# Output: dist/OpenCow-{version}.dmg

Cross-platform builds for Windows and Linux are supported via electron-builder configuration in package.json.


Contributing

We welcome contributions of all kinds — bug fixes, features, docs, and ideas.

# 1. Fork & clone
git clone https://github.com/<you>/opencow.git
cd opencow

# 2. Install & run
pnpm install && pnpm dev

# 3. Create a branch, make changes, submit a PR
git checkout -b feat/my-feature

Please read CONTRIBUTING.md for the full guide on code style, commit format, and review process.


Community


Star History


License

Apache-2.0 — completely free and open source. No paid tiers, no usage limits, no subscriptions.

Third-party SDKs: OpenCow integrates with third-party AI SDKs (e.g. @opencow-ai/opencow-agent-sdk, @openai/codex-sdk) that are subject to their own license terms. Please review each SDK's license before use.


Autonomous Agents for Every Team.

Collected info

  • 396 stars
  • 26 forks
  • Language: TypeScript
  • Source updated: 7/17/2026