> cowork-os --mode collaborative
workspace: local-first agents
Tip: approve risky shell actions before execution.
Hey CoWork, triage my release checklist
100% context left
Install the CLI
Works anywhere Node.js runs.
Free · Open-source · Everything app · Local-first · BYOK · No telemetry
ComparisonClaude Cowork or OpenClaw? Compare them with the local-first everything app.Open-source everything app
for AI agents.
Install CoWork OS on macOS, Windows, or Linux as an AI super app for chat, code, research, docs, spreadsheets, decks, and automations. Keep local-first approvals, visibility, multi-provider support, and a free open-source stack when comparing Claude Cowork or OpenClaw.
$npm install -g cowork-os@latest && cowork-os
Desktop app, npm install, or source build for your AI everything workspace.
Looking for a specific kind of agent platform?
MIT licensed
Open-source repo, releases, and documentation on GitHub.
macOS, Windows, Linux
Desktop installers, npm install, and self-hosted deployment paths.
Install and use in minutes
Download the app, use npm, or build from source with docs-backed setup.
Changelog and releases
Track release notes, version history, and project progress from canonical docs.
Product preview
Start from the app, then move into real work
The main workspace keeps suggested workflows, active context, recent runs, and the task composer in view.
Why people install CoWork OS
Built for real agent workflows — not experiments. Run agents locally with full visibility and control.
Run real agent workflows
Desktop, headless, remote, and channel-based execution. Four per-task modes: Autonomous, Collaborative, Multi-LLM, and Think With Me.
Stay in control
Approval-gated destructive actions, spending controls, and sandbox isolation. Your machine, your keys, your data.
See everything
Timeline events, outputs, files, decisions, and operator visibility — all in one place. No blind spots.
Choose how to get started with CoWork OS
Download the desktop app, install from your terminal, or connect with the open-source repo when you want full control of the runtime.
Let's build
Workspace app
Create a recurring standup from GitHub activity
+
Download the app
Desktop installers for local-first work.
I prepared the local dev workspace, checked docs routes, and staged the release workflow.
3 files readyReview ↗
package.json +12
README.md +28
Build from source
+
Build from source
Clone the repo and customize the stack.
Connect to GitHub
Use the repository for releases, docs, issues, and source-level review.
Get started with CoWork OS
SecurityApprovals and local-first controlsCoWork OS appDownload desktop installersTry in your terminalInstall with npm globallyOpen the repoClone, build, and inspect source
macOS first launch may require Privacy & Security → Open Anyway.
Windows SmartScreen may require More info → Run anyway.
Built in the open. Ready to inspect.
Repo-backed metrics, docs, releases, and install paths you can verify yourself.
30+ providers
Anthropic, OpenAI, Google, Ollama, AWS Bedrock, OpenRouter, and more.
Providers docs →
15 channels
WhatsApp, Telegram, Discord, Slack, Teams, Signal, iMessage, X, and more.
Channels docs →
139 skills
Built-in skills for code, docs, automation, infra, finance, and operations.
Features docs →
3200+ tests
Security-hardened quality coverage for production reliability.
Security guide →
Source of truth: public docs, GitHub releases, and the open repository.View source ↗
From request to controlled execution
A clear, approval-gated execution loop from request to output.
Step 1
Describe the outcome
Tell CoWork OS what you want. It creates an execution plan automatically.
Step 2
Let it plan and execute
Choose a mode — Autonomous, Collaborative, Multi-LLM, or Think With Me — and let the agent run.
Step 3
Watch the timeline and outputs
Track timeline events, shell output, generated files, and agent decisions in real time.
Step 4
Approve risky actions only
Destructive actions and spending remain explicitly approval-gated by design.
OpenClaw Alternative: CoWork OS
Respectful comparison for teams deciding between experimentation tooling and production operations.
| Positioning Lens | OpenClaw | CoWork OS |
|---|---|---|
| Category | Agent toolkit / experimentation layer | Security-hardened AI operating system |
| Primary use case | Build and iterate on agent workflows | Run agent workflows in production |
| Security posture | Varies by workflow and setup | Built-in approvals, guardrails, sandbox isolation, encrypted local storage |
| Runtime surfaces | CLI-centric control paths | Desktop app + headless daemon + 15-channel gateway |
| Data ownership | BYOK/self-hosting workflows | Local-first, BYOK, no telemetry, optional offline Ollama |
Alternatives & Use Cases
Explore focused pages targeting the highest-intent search topics around CoWork OS.
OpenClaw alternative
Production-focused comparison and migration framing.
Claude Cowork alternative
Free, open-source local alternative to hosted AI coworking tools.
personal AI assistant
Execution-first assistant model for real operational workflows.
agent operating system
Runtime, control, and context layers for team operations.
agentic systems
Deployment patterns for controlled, production agent execution.
use cases
Showcase pages, including a novel built end to end with the Novelist skill.
What makes CoWork OS different
Production differentiators from the README and docs, focused on reliability and operator visibility.
Mission Control
Kanban board, heartbeat monitoring, standup reports, and centralized orchestration.
Read docs →
Output Completion UX
High-signal completion toasts, file visibility cues, and artifact parity.
Read docs →
Active Context Sidebar
Live connectors and enabled skills are always visible with auto-refresh.
Read docs →
Usage Insights
Cost/token tracking, heatmaps, top skills, and pack-level analytics.
Read docs →
AI Playbook
Learns what worked from successful runs and feeds patterns into future tasks.
Read docs →
ChatGPT History Import
Import prior conversation history to avoid cold-start context loss.
Read docs →
Agent Runtime & Teams
Task-based execution with dynamic re-planning, multi-agent collaboration, and persistent teams. Watch the real-time timeline as the agent plans, executes, and produces artifacts.
- Four per-task modes: Autonomous, Collaborative, Multi-LLM, Think With Me
- Agent teams with shared checklists and multi-LLM synthesis
- High-signal completion UX with explicit output visibility
- Mission Control dashboard with Kanban board and heartbeat monitoring
Live Canvas & Build Mode
Agent-driven visual workspace for interactive content, data visualization, and iterative development workflows.
- Phased idea-to-prototype workflow (Concept → Plan → Scaffold → Iterate)
- Named checkpoints and revert support
- Interactive HTML/CSS/JS content rendering
- Iterative image annotation and data visualization
Works With Your Existing Stack
Connect CoWork OS to your messaging channels, providers, and infra surfaces without rebuilding your workflow.
Channels
WhatsAppTelegramDiscordSlackTeamsSignaliMessageX
Models
OpenAIAnthropicGoogleAWS BedrockOpenRouterOllama
Infrastructure
DockerE2BGitHubDomain APIsx402 Payments
Security-First Architecture
Top security score on ZeroLeaks with approvals, guardrails, sandbox isolation, and encrypted local storage. Includes 3200+ tests with 132+ security unit tests and 259+ WebSocket protocol tests.
Approval Workflows
Explicit consent required for destructive operations and spending actions.
Sandbox Isolation
macOS sandbox-exec, Docker isolation, or process-level sandboxing.
Local-First + BYOK
No telemetry middleman. Keys and data remain on your machine.
Deployment Modes
Choose the runtime form that matches your environment and operational requirements.
Install CoWork OS today
Free, open-source, and ready to run on your machine. Download the app, install via npm, or build from source.